You are on page 1of 67

NBER WORKING PAPER SERIES

RELATIONSHIP AND TRANSACTION LENDING IN A CRISIS


Patrick Bolton
Xavier Freixas
Leonardo Gambacorta
Paolo Emilio Mistrulli
Working Paper 19467
http://www.nber.org/papers/w19467
NATIONAL BUREAU OF ECONOMIC RESEARCH
1050 Massachusetts Avenue
Cambridge, MA 02138
September 2013
We would like to thank Claudio Borio, Mariassunta Giannetti, Roman Inderst, Natalya Martinova,
Lars Norden, Enrico Perotti, Joel Shapiro, Greg Udell and in particular two anonymous referees for
comments and suggestions. Support from 07 Ministerio de Ciencia e Innovacin, Generalitat de Catalunya,
Barcelona GSE, Ministerio de Economa y Competitividad) - ECO2008-03066, Banco de Espaa-Excelencia
en Educacin-"Intermediacin Financiera y Regulacin" is gratefully acknowledged. This study was
in part developed while Paolo Emilio Mistrulli was ESCB/IO expert in the Financial Research Division
at the European Central Bank. The views expressed herein are those of the authors and do not necessarily
reflect the views of the Bank of Italy, the Bank for International Settlements, or the National Bureau
of Economic Research.
NBER working papers are circulated for discussion and comment purposes. They have not been peer-
reviewed or been subject to the review by the NBER Board of Directors that accompanies official
NBER publications.
2013 by Patrick Bolton, Xavier Freixas, Leonardo Gambacorta, and Paolo Emilio Mistrulli. All
rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit
permission provided that full credit, including notice, is given to the source.
Relationship and Transaction Lending in a Crisis
Patrick Bolton, Xavier Freixas, Leonardo Gambacorta, and Paolo Emilio Mistrulli
NBER Working Paper No. 19467
September 2013
JEL No. G01,G21,G32
ABSTRACT
We study how relationship lending and transaction lending vary over the business cycle. We develop
a model in which relationship banks gather information on their borrowers, which allows them to provide
loans for profitable firms during a crisis. Due to the services they provide, operating costs of relationship-
banks are higher than those of transaction-banks. In our model, where relationship-banks compete
with transaction-banks, a key result is that relationship-banks charge a higher intermediation spread
in normal times, but offer continuation-lending at more favorable terms than transaction banks to profitable
firms in a crisis. Using detailed credit register information for Italian banks before and after the Lehman
Brothers' default, we are able to study how relationship and transaction-banks responded to the crisis
and we test existing theories of relationship banking. Our empirical analysis confirms the basic prediction
of the model that relationship banks charged a higher spread before the crisis, offered more favorable
continuation-lending terms in response to the crisis, and suffered fewer defaults, thus confirming the
informational advantage of relationship banking.
Patrick Bolton
Columbia Business School
804 Uris Hall
New York, NY 10027
and NBER
pb2208@columbia.edu
Xavier Freixas
Universitat Pompeu Fabra
Department of Economics and Business
Ramon Trias Fargas, 25-27
08005 Barcelona
Spain
xavier.freixas@upf.edu
Leonardo Gambacorta
Bank for International Settlements
Centralbanplatz, 2 4002 Basel
Switzerland
leonardo.gambacorta@bis.org
Paolo Emilio Mistrulli
Structural Economic Analysis Department
Bank of Italy
via Nazionale 91, 00184 Rome (IT)
paoloemilio.mistrulli@bancaditalia.it
1 Introduction
1
How do banks help their corporate borrowers through a crisis? Beyond pro-
viding loans to rms, commercial banks have long been thought to play a
larger role than simply screening loan applicants one transaction at a time.
By building a relationship with the rms they lend to, banks also play a
continuing role of managing rms nancial needs as they arise, whether in
response to new investment opportunities or to a crisis. What determines
whether a bank and a rm build a long-term relation, or whether they simply
engage in a market transaction? And, how do relationship and transaction
lending dier in a crisis? Our knowledge so far is still limited. To quote
Allen Berger, What we think we know about small versus large banks (...)
in small business lending may not be true and we know even less about them
during nancial crises.
2
We address these questions from both a theoretical
and empirical perspective.
Existing theories of relationship banking typically do not allow for aggre-
gate shocks and crises. Thus, we expand the relationship lending model of
Bolton and Freixas (2006) by introducing an aggregate shock along idiosyn-
cratic cash-ow risk for non-nancial corporations. In the expanded model
rms dier in their exposure to the aggregate shock and therefore may have
dierent demands for the nancial exibility provided by relationship bank-
ing. To be able to bring the model to the data we introduce a further critical
modication to the Bolton and Freixas model by allowing rms to borrow
from multiple banks on either a transaction or relationship basis.
3
1
We would like to thank Claudio Borio, Mariassunta Giannetti, Roman Inderst, Natalya
Martinova, Lars Norden, Enrico Perotti, Joel Shapiro, Greg Udell and in particular two
anonymous referees for comments and suggestions. The opinions expressed in this paper
are those of the authors only and do not necessarily reect those of the Bank of Italy or the
Bank for International Settlements. Support from 07 Ministerio de Ciencia e Innovacin,
Generalitat de Catalunya, Barcelona GSE, Ministerio de Economa y Competitividad) -
ECO2008-03066, Banco de Espaa-Excelencia en Educacin-Intermediacin Financiera
y Regulacin is gratefully acknowledged. This study was in part developed while Paolo
Emilio Mistrulli was ESCB/IO expert in the Financial Research Division at the European
Central Bank.
2
Keynote address, Indiana Review of Finance Conference, Dec. 2012.
3
The Bolton and Freixas (2006) model of relationship banking considers rms choice
of the optimal mix of nancing between a long-term banking relationship and funding
2
The main predictions from the theoretical analysis are: rst, that rms
will generally seek a mix of relationship lending (or, for short, 1banking)
and transaction lending (or 1banking); second, the rms relying on 1banks
are the ones that are more exposed to business-cycle risk and that have the
riskier cash ows. These rms are prepared to pay higher borrowing costs on
their relationship loans in normal times in order to secure better continuation
nancing terms in a crisis. Third, the rms relying on a banking relation are
better able to weather a crisis and are less likely to default than rms relying
only on transaction lending, even though the underlying cash ow risk of
rms borrowing from an 1bank is higher than that of rms relying only
on 1banking. Fourth, interest rates on 1loans are countercyclical: they
are higher than interest rates on 1loans in normal times and lower in crisis
times.
We test these predictions by looking at bank lending to rms in Italy be-
fore and after the Lehman Brothers default. Following Detragiache, Garella
and Guiso (2000), we use the extremely detailed credit registry information
on corporate lending by Italian banks, which allows us to track Italian rms
borrowing behavior before and after the crisis of 2008-09 at the individual
rm and bank level. The empirical analysis conrms the predictions of the
model. In particular, that relationship banks charged a higher spread before
the crisis, oered more favorable continuation-lending terms in response to
the crisis, and suered fewer defaults, thus conrming the informational and
nancial exibility advantage of relationship banking.
Our study is the rst to consider how relationship lending responds to
a crisis in a comprehensive way both from a theoretical and an empirical
perspective. Our sample covers loan contracts by a total of 179 Italian banks
to more than 72.000 rms over the time period ranging from 2007 to 2010,
with the collapse of Lehman Brothers marking the transition to the crisis.
The degree of detail of our data goes far beyond what has been available
in previous studies of relationship banking. For example, one of the most
important existing studies by Petersen and Rajan (1994) only has data on
rms balance sheets and on characteristics of their loans, without additional
through a corporate bond issue. Most rms in practice are too small to be able to tap
the corporate bond market, and the choice between issuing a corporate bond or borrowing
from a bank is not really relevant to them. However, as we know from Detragiache, Garella
and Guiso (2000) these rms do have a choice between multiple sources of bank lending
(see also Bolton and Scharfstein, 1996; Houston and James, 1996; Farinha and Santos,
2002).
3
specic information on the banks rms are borrowing from.
4
As a result
they cannot control for bank specic characteristics. We are able to do
so for both bank and rm characteristics, since we observe each bank-rm
relationship. More importantly, by focusing on multiple lender situations we
can run estimates with both bank and rm xed eects, thus controlling for
observable and unobservable supply and demand factors. We are therefore
able to precisely uncover the eects of bank-rm relationship characteristics
on lending. It turns out that our results dier signicantly depending on
whether we include or exclude these xed eects, revealing that the lack of
detailed information on each loan may lead to biases if, as one may expect, the
heterogeneity of banks (small, regional, large, mutuals,...) maps into dierent
lending behaviors that only bank xed eects can identify. Also, unlike the
vast majority of existing empirical studies, our database includes detailed
information on interest rates for each loan. This allows us to investigate
bank interest rate determination in good and bad times in a direct way,
without relying on any assumptions.
5
Overall, our study suggests that relationship banking plays an important
role in dampening the eects of negative shocks following a crisis. The rms
that rely on relationship banks are less likely to default on their loans and are
better able to withstand the crisis thanks to the more favorable continuation
lending terms they can get from 1banks. These ndings suggest that the
focus of Basel III on core capital and the introduction of countercyclical
capital buers could enhance the role of 1banks in crises and reduce the
risk of a major credit crunch especially for the rms that choose to rely on
4
They have a dummy variable taking the value 1 if the loan has been granted by a
bank and 0 if granted by another nancial institution, but they do not have information
on which bank has granted the loan and they do not have balance sheet information on
the bank.
5
In one related paper Gambacorta and Mistrulli (2013) investigate whether bank and
lender-borrower relationship characteristics had an impact on the transmission of the
Lehman default shock by analysing changes in bank lending rates over the period 2008:Q3-
2010:Q1. Bonaccorsi di Patti and Sette (2012) take a similar approach over the period
2007:Q2-2008:Q4, while Gobbi and Sette (2012) consider 2008:Q3-2009:Q3. Albertazzi
and Marchetti (2010) and De Mitri et al. (2010) complement the previous studies by
investigating the eect of the nancial crisis on |c:di:q qront/. In this paper, we focus
instead on the level of lending rates and the quantity of credit (instead than their respec-
tive changes). Moreover, we analyse the behaviour of relationship and transactional banks
by comparing bank prices and quantities both in normal times and in a crisis. Although
our results are not perfectly comparable, they are consistent with the above cited papers.
4
1banks.
Related Literature: Relationship banking can take dierent forms, and
most of the existing literature emphasizes benets from a long-term banking
relation to borrowers that are dierent from the nancial exiblity benets
that we model. The rst models on relationship banking portray the relation
between the bank and the rm in terms of an early phase during which
the bank acquires information about the borrower, and a later phase during
which it exploits its information monopoly position (Sharpe 1990). While
these rst-generation models provide an analytical framework describing how
the long-term relation between a bank and a rm might play out, they do
not consider a rms choice between transaction lending and relationship
banking, and which types of rms are likely to prefer one form of borrowing
over the other. They also do not allow for any rm bargaining power at
the default and renegotiation stage, as we do following Diamond and Rajan
(2000, 2001, and 2005).
The second-generation papers of relationship banking that consider this
question and that have been put to the data focus on three dierent and
interconnected roles for an 1bank: insurance, monitoring and screening.
A rst strand of studies focuses on the (implicit) insurance role of 1banks
against the risk of changes in future credit terms (Berger and Udell, 1992;
Berlin and Mester 1999); a second strand focuses on the monitoring role of
1banks (Holmstromand Tirole 1997, Boot and Thakor 2000, Hauswald and
Marquez 2000); and a third strand plays up the greater screening abilities of
new loan applications of 1banks due to their access to both hard and soft
information about the rm (Agarwal and Hauswald 2010, Puri et al. 2010).
Our theory is closest to a fourth strand which emphasizes the 1banks
ability to learn about changes in the borrowers creditworthiness, and to
adapt lending terms to the evolving circumstances the rm nds itself in
(Rajan, 1992 and Von Thadden, 1995). Interestingly, these four dierent
strands have somewhat dierent empirical predictions. Overall, our empirical
results suggest that only the predictions of the fourth strand of theories are
fully conrmed in our data. While all theories predict that 1banks have
lower loan delinquency rates than 1banks, only the fourth strand of theories
predicts that 1banks raise loan interest rates more than 1banks in crisis
times.
The structure of the paper is the following. In section 2 we describe the
theoretical model of 1banking and 1banking and in section 3 the com-
bination of the two forms of funding by rms. In section 4 we compare the
5
rms benets from pure 1banking with the ones of mixed nance, and the
implication for the capital buers the banks have to hold. In section 5 we
describe the database and we test the models predictions. Section 6 com-
pares our results with those derived by other types of theories of relationship
banking. The last section concludes.
2 The model
We consider the nancing choices of a rm that may be more or less ex-
posed to business-cycle risk. The rm may borrow from a bank oering
relationship-lending services, an 1bank, or from a bank oering only trans-
action services, a 1bank. As we explain in greater detail below, 1banks
have higher intermediation costs than 1banks, j
1
j
T
, because they have
to hold more equity capital against the expectation of more future roll-over
lending. We shall assume that the banking sector is competitive, at least
ex ante, before a rm is locked into a relationship with an 1bank. There-
fore, in equilibrium each bank just breaks even and makes zero supra-normal
prots. We consider in turn, 100% 1 bank lending, 100% 1bank lending,
and nally a combination of 1 and 1bank lending.
2.1 The Firms Investment and Financial Options
The rms manager-owners have no cash but have an investment project
that requires an initial outlay of 1 = 1 at date t = 0 to be obtained through
external funding. If the project is successful at time t = 1, it returns \
1
. If
it fails, it is either liquidated, in which case it produces \
1
at time t = 1,
or it is continued in which case the projects return depends on the rms
type, H or 1. For the sake of simplicity, we assume that the probability of
success of a rm is independent of its type. An Hrms expected second
period cash ow is \
1
, while it is zero for an 1rm. The probability that a
rm is successful at time t = 1 is observable, and the proportion of Hrms
is known. Moreover, both the probability of success and the proportion of
Hrms change with the business cycle, which we model simply as two
distinct states of the world: a good state for booms (o = G) and a bad state
for recessions (o = 1). Figure 1 illustrates the dierent possible returns of
the project depending on the banks decision to liquidate or to roll over the
6
unsuccessful rm at time t = 1.
6
We denote the rms probability of success at t = 1 as j
S
, with j
G
j
1

0. We further simplify our model by making the idiosyncratic high (\
1
) and
low (0) returns of rms at time t = 2 independent of the business cycle; only
the population of Hrms, which we denote by i
S
will be sensitive to the
business cycle. Finally, recession states (o = 1) occur with probability o
and boom states (o = G) occur with the complementary probability (1 o).
The prior probability (at time t = 0) that a rm is of type H is denoted
by i. This probability belief evolves to respectively i
1
in the recession state
and i
G
in the boom state at time t = 1, with i
1
< i
G
. The conditional
probability of a rm being of type H knowing it has defaulted in time t = 1
will be denoted by
i
(1 o)(1 j
G
)i
G
+o(1 j
1
)i
1
(1 j)
.
As in Bolton and Freixas (2006), we assume that the rms type is private
information at time t = 0 and that neither 1 nor 1 banks are able to identify
the rms type at t = 0. At time t = 1 however, 1banks are able to observe
the rms type perfectly by paying a monitoring cost : 0, while 1banks
continue to remain ignorant about the rms type (or future prospects).
Firms dier in the observable probability of success j = oj
1
+(1 o)j
G
.
For the sake of simplicity we take j
G
= j
1
+ and assume that j
G
is
uniformly distributed on the interval [. 1], so that j
1
is l [0. 1 ] and
j is l [(1 o). 1 o]. Note that for every j there is a unique pair
(j
1
. j
G
) so that all our variables are well dened.
Firms can choose to nance their project either through a transaction
bank or through a relationship bank (or a combination of transaction and re-
lationship loans). To keep the corporate nancing side of the model as simple
as possible, we do not allow rms to issue equity. The main distinguishing
features of the two forms of lending are the following:
6
A model with potentially innitely-lived rms subject to periodic cash-ow shocks
and that distinguishes between the value to the rm and to society of being identied as
an H-type, would be a better representation of actual phenomena. In a simplied way our
model can be reinterpreted so that the value of , takes already into account this long run
impact on the rms reputation. Still, a systematic analysis of intertemporal eects would
require tracking the balance sheets for both the rm and of two types of banks as state
variables of the respective value functions and would lead to an extremely complex model.
7
1. Transaction banking: a transaction loan species a gross repayment
:
T
(j) at t = 1. If the rm does not repay, the bank has the right to
liquidate the rm and obtains \
1
. But the bank can also oer to roll
over the rms debt against a promise to repay :
S
T
(j
S
) at time t = 2.
This promise :
S
T
(j
S
) must, of course, be lower than the rms expected
second period pledgeable cash ow, which is \
1
for an Hrm and
zero for an 1rm. Thus, if the transaction banks belief i
S
that it is
dealing with an Hrm is high enough, so that
:
S
T
(j
S
) i
S
\
1
.
the rm can continue to period t = 2 even when it is unable to repay
its debt :
T
(j) at t = 1. If the bank chooses not to roll over the rms
debt, it obtains the liquidation value of the rms assets \
1
at t = 1.
The market for transaction loans at time t = 1 is competitive and
since no bank has an informational advantage on the credit risk of the
rm the roll-over terms :
S
T
(j
S
) are set competitively. Consequently, if
gross interest rates are normalized to 1, competition in the 1banking
industry implies that
i
S
:
S
T
(j
S
) = :
T
(j). (1)
(when the project fails at time t = 1 the rm has no cash ow available
towards repayment of :
T
(j); it therefore must roll over the entire loan
to be able to continue to date t = 2).
For simplicity, we will assume that in the boom state an unsuccessful
rm will always be able to get a loan to roll over its debt :
T
(j):
:
T
(j)
i
G
\
1
. (2)
A sucient condition for this inequality to hold is that it is satised
for j
G
= .
7
By the same token, in a recession state rms with a high probability
of success will be able to roll over their debt :
T
(j) if i
1
is such that
:
T
(j)
i
1
\
1
. (3)
7
Note that the condition is not necessary as in equilibrium some rms with low j may
not be granted credit at time t = 0 anyway.
8
This will occur only for values of j
1
above some threshold j
1
for which
condition (3) holds with equality, a condition that, under our assump-
tions, is equivalent to j b j. where b j = j
1
+(1 o). In other words,
for low probabilities of success j < b j, an unsuccessful rm at t = 1
in the recession state will simply be liquidated, and the bank then re-
ceives \
1
, and for higher probabilities of success, j b j (or j
1
j
1
)
an unsuccessful rm at t = 1 in the recession state will be able to roll
over its debt. Figure 2 illustrates the dierent contingencies for the
case j
1
j
1
.
2. Relationship banking: Under relationship banking the bank incurs
a monitoring cost : 0 per unit of debt,
8
which allows the bank to
identify the type of the rm perfectly in period 1. A bank loan in period
0 species a repayment :
1
(j) in period 1 that has to compensate the
bank for its higher funding costs j
1
j
T
.
The higher cost of funding is due to the need of holding higher amounts
of capital that are required in anticipation of future roll-overs. It can
be shown, by an argument along the lines of Bolton and Freixas (2006),
that as the 1-banks are nancing riskier rms, even if, on average their
interest rates will cover the losses, they need additional capital. In ad-
dition 1-banks renance H-rms and they do so by supplying lending
to those rms that do not receive a roll-over from 1banks. As a con-
sequence, they also need more capital because of capital requirements
due to the expansion of lending to Hrms.
If the rm is unsuccessful at t = 1 the relationship bank will be able
to extend a loan to all the rms it has identied as Hrms and then
determines a second period repayment obligation of :
1
1
. As the bank
is the only one to know the rms type, there is a bilateral negotiation
over the terms :
1
1
between the rm and the bank. We let the rms
bargaining power be (1 ,) so that the outcome of this bargaining
process is :
1
1
= ,\
1
and the Hrms surplus from negotiations is
(1 ,)\
1
.
In sum, the basic dierence between transaction lending and relationship
8
Alternatively, the monitoring cost could be a xed cost per rm, and the cost would
be imposed on the proportion i of good rms in equilibrium. This alternative formulation
would not alter our results.
9
lending is that transaction banks have lower funding costs at time t = 0 but
at time t = 1 the rms debt may be rolled over at dilutive terms if the
transaction banks beliefs that it is facing an Hrm i
1
are too pessimistic.
Moreover, the riskiest rms with j < b j will not be able to roll over their
debts with a transaction bank in the recession state. Relationship banking
instead oers higher cost loans initially against greater roll-over security but
only for Hrms.
3 Equilibrium Funding
Our set up allow us to determine the structure of funding and interest rates
at time t = 1 and t = 2 under alternative combinations of transaction and
relationship loans. We will consider successively the cases of pure transaction
loans, pure relationship loans, and a combination of the two types of loans.
We assume for simplicity that the intermediation cost of dealing with a bank,
whether 1bank or 1bank is entirely capitalized in period 0 and reected
in the respective costs of funds, j
T
and j
1
. We will assume as in Bolton and
Freixas (2000, 2006) that Hrms move rst and 1rms second. The latter
have no choice but to imitate Hrms by pooling with them, for otherwise
they would perfectly reveal their type and receive no funding.
Transaction Banking: Suppose that the rmfunds itself entirely through
transaction loans. Then the following proposition characterizes equilibrium
interest rates and funding under transaction loans.
Proposition 1: Under 1banking, rms characterized by j b j are
never liquidated and pay an interest rate
:
S
T
=
1
i
S
on their rolled over loans.
For rms with j < b j there is no loan roll-over in recessions, and the
roll-over of debts in booms is granted at the equilibrium repayment promise:
:
G
T
=
1
i
G
.
10
The equilibrium lending terms in period 0 are then:
:
T
(j) = 1 +j
T
for j b j. (4)
:
T
(j) =
1 +j
T
o(1 j
1
)\
1
oj
1
+ 1 o
for j < b j.
Proof: See Appendix A.
Relationship Banking: Consider now the other polar case of exclu-
sive lending from an 1bank. The equilibrium interest rates and funding
dynamics are then given in the following proposition.
Proposition 2: Under relationship-banking there is always a debt roll-
over for Hrms at equilibrium terms
:
1
1
= ,\
1
.
The equilibrium repayment terms in period 0 are then given by:
:
1
(j) =
1 +j
1
(1 j)[(1 i)\
1
+i(1 :),\
1
]
oj
1
+ (1 o)j
G
. (5)
Proof: See Appendix A.
Combining 1 and 1banking: In the previous two cases of either
pure 1 banking or pure 1banking the structure of lending is independent
of the rms type. When we turn to the combination of 1 and 1banking,
the rms choice might signal their type. As mentioned this implies that the
1rms will have no choice but to mimick the Hrms.
Given that transaction loans are less costly (j
T
< j
1
) it makes sense for
a rm to rely as much as possible on lending by 1banks. However, there is
a limit on how much a rm can borrow from 1banks, if it wants to be able
to rely on the more ecient debt restructuring services of 1banks. The
limit comes from the existence of a debt overhang problem if the rms are
overindebt with 1banks.
To see this, let 1
1
and 1
T
denote the loans granted by respectively
1banks and by 1banks at t = 0, with 1
1
+1
T
= 1. Also, let :
1T
1
and :
1T
T
denote the corresponding repayment terms under each type of loan. When a
rm has multiple loans an immediate question arises: what is the seniority
11
structure of these loans? As is common in multiple bank lender situations,
we shall assume that 1bank loans and 1bank loans are pari passu in the
event of default. Under this assumption, the following proposition holds:
Proposition 3: The optimal loan structure for Hrms is to maxi-
mize the amount of transactional loans subject to satisfying the relationship
lenders incentive to roll over the loans at t = 1.
The rm borrows:
1
T
=
(j + (1 j)i)

,\
1
(1 :) \
1

1 +j
T
\
1
. (6)
in the form of a transaction loan, and (11
T
) from an 1bank at t = 0
at the following lending terms:
:
1T
T
=
(1 +j
T
) (1 j)(1 i)\
1
j + (1 j)i
. (7)
and,
:
1T
1
=
1
j

(1 +j
1
)
(1 j)\
1
(1 1
T
)

. (8)
At time t = 1 both transaction and relationship-loans issued by H rms
are rolled over by the 1-bank. Neither loan issued by an 1 rm is rolled
over.
Proof: See Appendix A.
As intuition suggests: i) pure relationship lending is dominated under our
assumptions; and ii) if the bank has access to securitization or other forms of
funding to obtain funds on the same terms as 1banks, then it can combine
the two.
Note nally that, as 1loans are less expensive, a relatively safe rm
(with a high j) may still be better o borrowing only from 1banks and
taking the risk that with a small probability it wont be restructured in
bad times. We turn to the choice of optimal mixed borrowing versus 100%
1nancing in the next section.
12
4 Optimal funding choice
When would a rm choose mixed nancing over 100% 1nancing? To
answer this question we need to consider the net benet to an Hrm from
choosing a combination of 1 and 1bank borrowing over 100% 1bank
borrowing. We will make the following plausible simplifying assumptions in
order to focus on the most interesting parameter region and limit the number
of dierent cases to consider:
Assumption A1: Both (j
1
j
T
) and : are small enough.
Assumption A2: ,\
1
\
1
is not too large so that it satises:
,\
1
\
1
< min
(
(1 +j
T
)[
(10)
i

+
0
i

1]
(1 [(1 o)i
G
+oi
1
])
.
o(1 j
1
)(\
1
\
1
)
(1 j)(1 i)
)
These two conditions essentially guarantee that relationship banking has
an advantage over transaction banking. For this to be true, it must be the
case that: First, the intermediation cost of relationship banks is not too large
relative to that of transaction banks. Assumption A1 guarantees that this is
the case. Second, the cost of rolling over a loan with the 1bank should not
be too high. This means that the 1bank should have a bounded ex post
information monopoly power. This is guaranteed by assumption A2.
To simplify notation and obtain relatively simple analytical expressions,
we shall also assume that \
1

(j)
i

. The last inequality further implies


that \
1

(j)
i

. as i
G
i
1
. so that the rms debts will be rolled over by
the 1bank in both boom and bust states of nature. Note that when this is
the case the transaction loan is perfectly safe, so that :
T
(j) = 1 + j
T
. as in
equation (4).
We denote by (j) =
T
(j)
1T
(j) the dierence in expected payos
for an Hrm from choosing 100% 1nancing over choosing a combination
of 1 and 1loans and establish the following proposition.
Proposition 4: Under assumptions A1 and A2, the equilibrium funding
in the economy will correspond to one of the three following congurations:
1. (j
min
) ((1 o)) 0: monitoring costs are excessively high
and all rms prefer 100% transactional banking.
13
2. (j
max
) (1o) 0 and (j
min
) ((1o)) < 0: Safe
rms choose pure 1banking and riskier rms choose a combination
of 1banking and 1-banking.
3. (j
max
) (1 o) < 0: all rms choose a combination of
1banking and 1banking.
Proof: See Appendix A.
We are primarily interested in the second case, where we have coexistence
of 100% 1banking by the safest rms along with other rms combining
1Banking and 1banking. Notice, that under assumptions A1 and A2, it
is possible to write
((1 o)) = (j
1
j
T
)(1 1

T
) + (1 (1 o))

i:,\
1

+o(1 +j
T
) + (1 (1 o))(1 i)(,\
1
\
1
)
(1 +j
T
)[
(1 o)(1 )
i
G
+
o
i
1
] (9)
and
(1 o) = (1 +j
1
) (j
1
j
T
)1

T
o

i(1 :),\
1
+ (1 i)\
1

(1 o)(1 +j
T
) +o,\
1
(1 +j
T
)[
o
i
1
]. (10)
Under assumption A1 (j
1
j
T
) and : are small, so that a sucient
condition to obtain (1 o) 0 is to have o suciently close to zero.
Indeed, then we have:
(1 o) (j
1
j
T
)(1 1

T
) 0
To summarize the testable hypotheses coming from our theoretical model
are the following:
14
1 1banks are better able than 1banks at learning rms type. In a
crisis, the rate of default on rms nanced through transaction loans
will be higher than the rate on rms nanced by 1banks.
2 1banks charge higher lending rates in good times on the loans they
roll over, but in bad times they lower rates to help their best clients
through the crisis. 1banks increase their supply of lending (relative
to 1banks) in bad times.
3 It exists a critical threshold for the probability of success in bad time j
1
such that for any j
1
j
1
rms prefer pure transactional banking and
for any j
1
< j
1
rms prefer to combine the maximum of transactional
banking and the minimal amount of relationship banking.
4 Banks need a capital buer to be used in order to preserve the lending
relationship in bad times. This implies that the capital buer of an
1bank (which makes additional loans to good rms in distress) will
have to be higher than the one of a 1bank. This is consistent with
1banks quoting higher interest rates in normal times.
5 Data and empirical ndings
We now turn to the empirical investigation of relationship banking over the
business cycle. A key test we are interested in is whether 1banks charge
higher lending rates in good times and lower rates in bad times to help their
best clients through the crisis, and, similarly, whether 1banks oer cheaper
loans in good times but roll over fewer loans in bad times. Another related
prediction from our theoretical analysis we test is whether we observe lower
delinquency rates in bad times for 1banks that roll over their loans. To
test these predictions, we proceed in two steps. First we analyze how rms
default probability in bad times is inuenced by the fact that the loan is
granted by an 1bank or a 1bank. Second, we analyze (and compare)
lending and bank interest rate setting in good times and bad times. Our
dataset comes from the Italian Credit Register (CR) maintained by the Bank
of Italy and other sources.
The rst challenge we face is to identify two separate periods that repre-
sent the two states of the world in the model. Our approach is to distinguish
and compare bank-rm relationships prior to and after the Lehman Brothers
15
default (in September 2008), the event typically used to mark the beginning
of the global nancial crisis in other studies (e.g. Schularick and Taylor,
2011).
Our unique dataset covers a signicant sample of Italian banks and rms.
There are at least four advantages in focusing on Italy. First, from Italys
perspective the global nancial crisis was largely an unexpected (exogenous)
event, which had a sizable impact especially on small and medium-sized
rms that are highly dependent on bank nancing. Second, although Italian
banks have been hit by the nancial crisis, systemic stability has not been
endangered and government intervention has been negligible in comparison
to other countries (Panetta et al. 2009). Third, multiple lending is a long-
standing characteristic of the bank-rm relationship in Italy (Foglia et al.
1998, Detragiache et al. 2000). And fourth, the detailed data available for
Italy allow us to test the main hypotheses of the theoretical model without
making strong assumptions. Mainly, the availability of data at the bank-rm
level on both quantity and prices allows us to overcome major identication
problems encountered by the existing bank-lending-channel literature in dis-
entangling loan demand from loan supply shifts (see e.g. Kashyap and Stein,
1995; 2000).
The visual inspection of bank lending and interest rate dynamics in Figure
3 helps us to select two periods that can be considered good and bad times:
We select the second quarter of 2007 as a good time-period when lending
reached a peak and the interest rate spread applied on credit lines levelled to
a minimum (see the green circles in panels (a) and (b) of Figure 3). We take
the bad time-period to be the rst quarter of 2010, which is characterized by
a contraction in bank lending to rms and a very high intermediation spread
(see the red circles in panels (a) and (b) of Figure 3). The selection of these
two time-periods is also consistent with alternative indicators such as real
GDP and stock market capitalization (see panel (c) in Figure 3). We do not
consider the time-period beyond 2011, as it is aected by the eects of the
European Sovereign debt crisis.
Our second challenge is to distinguish between 1banks and 1banks.
One of the distinguishing characteristics of 1banks in our theory and other
relationship-banking models (Boot 2000; Berger and Udell 2006) is that
1banks gather information about the rms they lend to on an ongoing
basis, to be able to provide the exible nancing their client rms value.
Thus, the measure of relationship banking we focus on is the informational
16
distance between lenders and borrowers.
9
The empirical banking literature
has established that greater geographical distance between a bank and a rm
aects the ability of the bank to gather soft information (that is, information
that is dicult to codify), which in turn undermines the banks ability to
act as a relationship lender (see Berger et al. 2005, Agarwal and Hauswald
2010).
The theoretical banking literature argues that greater geographical dis-
tance plausibly increases monitoring costs and diminishes the ability of banks
to gather and communicate soft information. In particular, it is argued that
loan ocers who are typically charged with gathering this kind of informa-
tion and to pass it through the banks hierarchical layers will face higher
costs.
10
Stein (2002) has shown that when the production of soft informa-
tion is decentralized, incentives to gather it crucially depends on the ability
of the agent to convey information to the principal. Cremer, Garicano and
Prat (2007) have also argued that distance may aect the transmission of
information within banks (i.e. the ability of branch loan ocers to harden
soft information), since bank headquarters may be less able to interpret the
information they receive from distant branch loan ocers than from closer
ones. They show that there is a trade-o between the eciency of commu-
nication and the breadth of activities covered by an organization, so that
communication is more dicult when headquarters and branches are farther
apart.
We therefore divide 1banks and 1banks by the geographical distance
between the lending banks headquarters and rm headquarters, which we
take as a simple proxy for informational distance.
11
More precisely, we in-
troduce two dummy variables: for an 1bank, the rst variable is equal
to 1 if rm / is headquartered in the same province where bank , has its
headquarters; for a 1bank, the second variable is equal to 1 when the rst
variable (for the 1bank) is equal to 0. This way a bank can act both as
9
There is not a clear consensus in the literature on the way relationship characteris-
tics are identied. In Appendix C, we have checked the robustness of the results using
alternative measures for relationship lending.
10
Soft information it is gathered through repeated interaction with the borrower and
then it requires proximity. Banks in order to save on transportation costs delegate the
production of soft information to branch loan ocers since they are those within bank
organizations which are the closest to borrowers. Alternatively, one can consider the
geographical distance between bank branches and rms headquarters. However, Degryse
and Ongena (2005) nd that this measure has little relation to informational asymmetries.
11
Accordingly, branches of foreign banks are treated as Tbanks.
17
an 1bank for a rm headquartered in the same province and as a 1bank
for rms that are far away.
Our third challenge is to measure credit risk and to distinguish this mea-
sure from asymmetric information. One basic assumption of our theoretical
framework is that ex ante all banks know that some rms are more risky
than others. The objective measure of a rms credit risk shared ex ante by
all banks is given by (1 j) in our model and by a rms 2score in our
empirical analysis. The 2score, thus, is our measure of a rms ex-ante
probability of default. The 2scores can be mapped into four levels of risk:
1) safe; 2) solvent; 3) vulnerable; and 4) risky. Measuring asymmetric infor-
mation and the role of relationship banks in gathering additional information
is more complex. Obviously, no contemporaneous variable could possibly re-
ect soft information that is private to the rm and the relationship bank.
Consequently, it is only ex post that a variable could reect the skills of re-
lationship banks in renancing the good rms and liquidating the bad ones.
Relationship banks superior soft information must imply that rms who are
able to roll over their loans from a relationship bank must on average have
lower rates of default. This is why in order to distinguish Hrms from
1rms we look at the realization of defaults.
Table 1 gives some basic information on the dataset after having dropped
outliers (for more information see the data Appendix B). The table is di-
vided horizontally into three panels: i) all rms, ii) Hrms (i.e. rms that
have not defaulted during the nancial crisis) and iii) 1rms (i.e. rms
that have defaulted on at least one of their loans). In the rows we divide
bank-rm relationships into: i) pure relationship lending: rms which have
business relationship with 1banks only; ii) mixed banking relationship:
rms which have business relationships with both 1banks and 1banks;
iii) pure transactional lending: rms which have business relationship with
1banks only.
Several clear patterns emerge. First, cases where rms have only relation-
ships with 1banks (10% of the cases) or 1banks (44% of the cases) are
numerous but the majority of rms borrow from both kinds of banks (46%).
Second, the percentage of defaulted rms that received lending only from
1banks is relatively high (64% of the total). Third, in the case of pure
1banking, or combined 11 banking, rms experience a lower increase
in the spread in the crisis. Fourth, 1banking is associated with a higher
level of bank equity-capital ratios, so that 1banks have a buer against
contingencies in bad times. Their equity-capital slack depends on the busi-
18
ness cycle and is depleted in bad times. Interestingly, the size of the average
1bank is four times that of the average 1bank (100 vs 25 billions). This
is in line with Stein (2002) who points out that the internal management
problem of very large intermediaries may induce these banks to rely solely
on hard information in order to align the incentives of the local managers
with headquarters.
These patterns are broadly in line with the predictions of our theoretical
model. However, these ndings can only be suggestive as the bank-lending
relationship is inuenced not only by rms types but also by other factors
(sectors of a rms activity, the rms age and location, bank-specic charac-
teristics, etc.), which we have not yet controlled for in the descriptive sample
statistics reported in Table 1.
5.1 Hypothesis 1: 1banks have better information
than 1banks about rms credit risk
To verify whether 1banks are better able to learn rms types than 1banks
we look at the relationship between the probability that a rm / defaults
and the rms relative share of transactional vs relationship nancing. If
1banks have superior information than 1banks in a crisis then the rate
of default of rms nanced through transaction loans will be higher than for
rms nanced by 1banks.
Our baseline cross-sectional equation estimates the marginal probability
of default of rm / in the six quarters that follow the Lehman Brothers
collapse (2008:q3-2010:q1) as a function of the share of loans of rm / from a
1bank in 2008:q2. In particular, we estimate the following marginal probit
model:
`1(Firm ks default=1) = c + +:(1 :/c:c)
I
+
I
(11)
where c and are, respectively, vectors of bank and industry-xed eects
and (1 :/c:c)
I
is the pre-crisis proportion of transactional loans (in value)
for rm /. The results reported in Table 2 indicate that a rm with more
1bank loans has a higher probability of default.
12
This marginal eect
12
In principle the reliability of this test may be biased by the possible presence of
evergreening, a practice aimed at postponing the reporting of losses on the balance
sheet. Albertazzi and Marchetti (2010) nd some evidence of evergreening practices
in Italy in the period 2008:Q3-2009:Q1, although limited to small banks. We think that
19
increases with the share of 1bank nancing and reaches a maximum of
around 0.3% when (1 :/c:c)
I
is equal to 1. This eect is not only sta-
tistically signicant but also important from an economic point of view, as
the average default rate for the whole sample in the period of investigation
was around 1%. This nding is robust to enriching the set of controls with
additional rm-specic characteristics (see panel II in Table 2) or to calcu-
lating the proportion of transactional loans in terms of the number of banks
that nance rm / instead of by the size of outstanding loans (see panel III
in Table 2).
5.2 Hypotheses 2: 2a) 1banks charge higher rates
in good times and lower rates in bad times. 2b)
1banks increase their supply of lending (relative
to 1banks) in bad times.
In the second step of our analysis we investigate bank lending and price set-
ting over the business cycle. Our focus on multiple lending relations at the
rm level allows us to solve potential identication issues, by including both
bank and rmxed eects in the econometric model. In particular, the inclu-
sion of xed eects allows us to control for all (observable and unobservable)
time-invariant bank and borrower characteristics, and to identify in a precise
way the eects of bank-rm relationships on the interest rate charged and
the loan amount.
We estimate two cross-sectional equations: one for the interest rate (:
),I
)
applied by bank , on the credit line of rm /, and the other for the logarithm
evergreening is less of a concern in our case for three reasons.
First, evergreening is a process that by nature cannot postpone the reporting of losses
for a too long time. In our paper, we consider the period 2008:Q3-2010:Q1 that is 18
months after Lehmanns default and therefore there is a higher probability that banks
have reported losses.
Second, there is no theoretical background to argue that evergreening can explain the
dierence we document between T-banks and 1-banks. Both kinds of banks may have
a similar incentive to postpone the reporting of losses to temporarily inate stock prices
and protability.
Third, in the case that 1-banks have more incentive to evergreen loans the denition of
default used in the paper limits the problem. In particular, we consider a rm as in default
when at least one of the loans extended is reported to the credit register as a defaulted
one (the ag is up when at least one bank reports the client as bad). This means that
a 1-bank cannot eectively postpone the loss simply because a T-banks will report it.
20
of outstanding loans by bank , in real terms (1
),I
) on total credit lines to
rm /:
:
),I
= i +, +1-bank
),I
+
),I
. (12)
1
),I
= o +c +j1-bank
),I
+
),I
. (13)
where i and o are bank-xed eects and , and c are rm-xed eects.
13
Both
equations are estimated over good times (2007:q2) and bad times (2010:q1).
The results are reported in Table 3.
14
In line with the predictions of the
model, the coecients show that 1banks (compared to 1banks) provide
loans at a cheaper rate in good times and at a higher rate in bad times (see
columns I and II). The dierence between the two coecients
1

G
=
.123 (.081) = .204

is statistically signicant. As for loan quantities,


other things being equal, 1banks always provide on average a lower amount
of lending, especially in bad times (see columns III and IV). In this case
the dierence j
G
j
1
= .313 (.275) = 0.038

indicates that, other


things being equal, 1banks increase their supply of loans in bad times. In
particular, they supply 4% more loans relatively to 1banks.
13
It is worth stressing that the analysis of interest rates applied on credit lines is partic-
ularly useful for our purposes for two reasons. First, these loans are highly standardized
among banks and therefore comparing the cost of credit among rms is not aected by
unobservable (to the econometrician) loan-contract-specic covenants. Second, overdraft
facilities are loans granted neither for some specic purpose, as is the case for mortgages,
nor on the basis of a specic transaction, as is the case for advances against trade credit
receivables. As a consequence, according to Berger and Udell (1995) the pricing of these
loans is highly associated with the borrower-lender relationship, thus providing us with a
better tool for testing the role of lending relationships in bank interest rate setting.
14
Following Albertazzi and Marchetti (2010) and Hale and Santos (2009) we cluster
standard errors (-

) at the rm level in those regressions that include bank xed eects.


Vice versa in those regressions that include specic rm xed eects (but no bank xed
eects) we cluster standard errors at the bank group level. In this way we are able to
control for the fact that, due to the presence of an internal capital market, probably
nancial conditions of each bank in the group is not independent of one another. For a
general discussion on dierent approaches used to estimating standard errors in nance
panel data sets, see Petersen (2009).
21
5.3 Hypothesis 3: Safe rms prefer transactional lend-
ing; other rms prefer to combine transactional
and relationship banking.
A key prediction of our model is that rms with low underlying cash-ow
risk (those with a probability of success in bad times that is greater than j
1
)
prefer pure transaction banking, while those with higher cash-ow risk (with
j
1
j
1
) prefer to combine transaction and relationship banking. To test
this prediction we will look for a 2score relation such that rms with a low
2 score reveal their preference for pure transactional banking and those with
a high 2 score reveal their preference for combined 1 and 1banking. To
this end, equations (12) and (13) are further enriched with interaction terms
between bank-types and the 2score in order to explore whether 1banks
and 1banks behave dierently with respect to borrowers with a dierent
degree of risk:
:
),I
= i+o+(1bank)
),I
+
Z
(1bank)
),I
2+j
Z
(1bank)
),I
2+A+
),I
(14)
1
),I
= o+o+j1bank
),I
+j
Z
(1bank)
),I
2+
Z
(1bank)
),I
2+A+
),I
(15)
In the above equations we can include only bank xed eects, as the
interaction terms between bank type and 2scores (a linear combination
that is invariant for each rm) prevent us from including rm-xed eects.
For this reason we also enrich the set of controls by including a complete set of
industry-province dummies (o) and a vector A with a number of rm-specic
characteristics. In particular A now contains:
a dummy USGR that takes the value of 1 for those rms that have
used their credit lines for an amount greater than the value granted by
the bank, and zero elsewhere;
a dummy that takes the value of 1 if the rm is a limited liability
corporation, and zero elsewhere (LTD);
a dummy that takes the value of 1 for rms with less than 20 employees
(SMALL_FIRM), and zero elsewhere; this dummy aims to control for
the fact that small rms do not issue bonds as larger rms may do;
22
the length of the borrowers credit history (CREDIT HISTORY) mea-
sured by the number of years elapsed since the rst time a borrower
was reported to the Credit Register. This variable tells us how much
information has been shared among lenders through the Credit Register
over time and is a proxy for rms reputation acquisition.
The results are reported in Table 4. Firms that use their credit lines
for an amount greater than the value granted by the bank have to pay a
higher spread that increases in bad times. Repeated interaction with the
banking system also has an eect on bank interest rate setting and loan
supply. The variable CREDIT_HISTORY, representing the number of years
elapsed since the rst time a borrower was reported to the Credit Register, is
negatively (positively) correlated with rates applied to credit lines (amount
of outstanding loans). Firms organized as limited liability corporations (LTD
corporations) are less opaque and pay a lower spread. Other things being
equals, LTD corporations also need less bank lending because they have
access to other sources of funds.
The graphical representation of the interaction terms between bank-types
and the 2score is reported in Figure 4. The upper panels (a) and (b)
describes the eects on the interest rate, the bottom panels (c) and (d) those
on the logarithm of real loans. The graphs on the left illustrate the pre-
Lehman period, while those on the right represent the post-Lehman period.
In each graph the horizontal axis reports the 2score, where 2 goes from
1 (safe rm) to 4 (risky rm). Transaction banking (1banks) is indicated
with a dotted line and relationship banking (1banks) with a solid line.
The visual inspection of all graphs shows that both interest rates and
loan size are positively correlated with the 2score. The positive correlation
between risk and bank nancing probably reects the fact that risky rms
have a limited access to market nancing. As one would expect, the interest
rate increases with credit risk.
In line with the predictions of our model, the cost of credit of transactional
lending is always lower than relationship banking in good times: the dotted
line is always below the solid one for all 2scores (see panel (a) of Figure
4). This pattern is reversed in bad times (panel (b)) when banks with a
strong lending relationship (1banks) oer lower rates to :i:/ rms (those
with a 2score greater than 1). And, as predicted by our model, it is always
cheaper for safe rms to use transactional banking because they obtain always
a lower rate from 1banks. Moreover the two bottom panels of Figure 4
23
highlight that the roll-over eects of 1banks on lending is mostly present
for risky rms, while :c,c rms always obtain a greater level of nancing
from 1banks whether in good or bad times (the dotted line is always above
the solid line for 2 = 1 both in panel (c) and panel (d)).
5.4 Hypothesis 4: 1banks need capital buers to be
used to preserve the lending relationship in bad
times
A nal prediction of our model is that 1banks have a higher equity capital
buer than 1banks in good times so as to support the lending relationship
in bad times. To test this prediction we focus on bank capital endowments:
Since 1banks have a lower incentive of making additional loans to rms
in distress in bad times, we should observe that these banks hold a lower
equity-capital buer against contingencies relative to 1banks prior to the
crisis. In particular, we estimate the following cross-sectional equation on
our sample of 179 banks:
C1
)
= . +t(1 :/c:c)
)
+1 +2 +
)
. (16)
where the dependent variable C1
)
is the regulatory capital-to-risk-weighted-
assets of bank j in 2008:q2 (prior to the Lehman collapse), the variable
(1 :/c:c)
)
is the proportion of transaction loans (in value) for bank j, . is
a set of bank-zone dummies, 1 a set of bank-specic controls, and 2 a set of
bank credit portfolio-specic controls. Bank specic characteristics include
not only bank size and liquidity ratio (liquid assets over total assets), but also
the retail ratio between deposits and total bank funding (excluding capital).
All explanatory variables are taken in 2008:q1 in order to mitigate endogene-
ity problems. The results reported in Table 6 indicate that, indipendently
of the model specication chosen, a pure 1bank that has a credit portfolio
composed exclusively of transaction loans (1 :/c:c
)
= 1) have a capi-
tal buer more than 3 percentage points lower than a pure 1bank, whose
portfolio is composed exclusively of relationship loans (1 :/c:c
)
= 0).
Finally, we also test the eects of bank capital endowments on interest
rates and lending. We thus include in our baseline equations (12) and (13),
the regulatory capital-to-risk weighted assets ratio (C1, lagged one period
to mitigate endogeneity problems), a set of bank-zone dummies (.), and other
bank-specic controls (1 ):
24
:
),I
= , +. +(1 bank)
),I
+iC1
)
+1 +
),I
(17)
1
),I
= c +. +j(1 bank)
),I
+`C1
)
+1 +
),I
(18)
The vector 1 contains in particular the dummy lo G1, described above,
and:
a dummy for mutual banks (`l1l1), which are subject to a special
regulatory regime (Angelini et al., 1998);
a dummy equal to 1 if a bank belongs to a group and 0 elsewhere;
a dummy equal to 1 if a bank has received government assistance and
0 elsewhere.
The results reported in Table 5 indicate that banks with larger capital
ratios are better able to protect the lending relationship with their clients.
Well-capitalized banks have a higher capacity to insulate their credit portfolio
from the eects of an economic downturn by granting a higher amount of
lending at a lower interest rate. To get a sense of the economic impact of
the above-mentioned results, during a downturn a bank with a capital ratio 5
percentage points greater with respect to another one supplies 5% more loans
at an interest rate 20 basis points lower. This result on the eects of bank
capital is in line with the bank lending channel literature which indicates that
well-capitalized banks are better able to protect their clients in the event of a
monetary policy shock (Kishan and Opiela, 2000; Gambacorta and Mistrulli,
2004).
Interestingly, the positive eect of bank capital in protecting the lending
relationship is more important for 1banks than for 1banks. This can be
tested by replacing iC1 in equation (17) with

T
(1 /c:/) C1 +
1
(1 /c:/) C1
and `C1 in equation (18) with
`
T
(1 /c:/) C1 +`
1
(1 /c:/) C1.
In particular, the coecients
T
and
1
take the values of -0.054*** (s.e.
0.008) and -0.038*** (s.e. 0.010), respectively, and are statistically dierent.
25
A similar result is obtained in the lending equation, where `
T
and `
1
take
the values of -0.025*** (s.e. 0.005) and -0.006* (s.e. 0.003) respectively,
and are statistically dierent from one another. This means that during
a downturn an 1bank (resp. 1bank) with a capital ratio 5 percentage
points greater than another 1bank supplies 12% more loans at an interest
rate 27 basis points lower (resp. 3% more loans and 19 basis points lower
rates for a 1bank).
5.5 Robustness checks
We have checked the robustness of our results in several ways.
15
We have:
(1) included in the baseline regressions an additional measure of relationship
banking, namely a dummy for the main bank; (2) tested for alternative
denitions of the relationship dummy 1bank; (3) considered all foreign
banks as 1banks; (4) estimated equations on a subset of new rms. In
all cases results are very similar.
One distinctive feature of our dataset is that, by focusing on multiple
lending, we can run estimates with both bank and rm xed eects, thus
controlling for observable and unobservable supply and demand factors. In
this way we are able to clearly identify the eects of bank-rm relationship
characteristics on lending, not biased by the omission of some variables that
may aect credit conditions. To highlight this point we have re-run all the
models without bank and rm xed eects. The new set of results (not
reported for the sake of brevity but available from the authors upon request)
indicates that 1bank coecients are often dierent and may even change
sign in one third of the cases. In particular, when we do not introduce
xed-eects, 1banks are shown to supply relatively more lending but at
higher prices. This is an important observation, as it clearly shows that not
controlling for all unobservable bank and rm characteristics biases results;
in particular, the benets of relationship lending tend to be overestimated
on prices and underestimated on quantities.
The bias can be seen by comparing Figure 4 (illustrating the results of
the models with xed eects reported in Table 4) with Figure 5 (illustrating
results of the same models but without xed eects). In Figure 5 the dot-
ted line for 1banks moves upwards relative to Figure 4. In other words,
1banks supply relatively more loans and charge higher interest rates when
15
For more details see Appendix C.
26
xed eects are dropped compared with the case when they are added to the
estimated equation.
This last nding is consistent with the way we identify transaction and
relationship banking, as one would expect to see this sign for the bias if
1banks are better able than 1banks to gather soft information, and con-
sequently better able to discriminate good and bad borrowers. Indeed, con-
sider two rms that have the same 2score but one is, in reality, riskier than
the other. According to our model, 1banks are better able to discriminate
between the two while 1banks are not. This happens because 1banks
use only hard information (incorporated in the 2scores) while 1banks
rely on both hard and soft information. If this is true then the riskiest rm
would prefer to ask 1banks for a loan since these banks are less able to
distinguish good from bad borrowers. As a consequence, the riskiest rm
will in theory get better price conditions and a larger amount of credit com-
pared to the situation where 1banks are able to evaluate rms credit risk
correctly. Of course, 1banks are perfectly aware that the bank-rm match-
ing is not random and that their applicant pool is on average riskier than
that of 1banks. As a consequence, 1banks will charge higher interest
rates, anticipating that they will tend to make more mistakes in their loan
restructuring choices compared to 1banks. In other words, 1banks know
that they will tend to lend too much and this is exactly what we observe
in Figure 5, where we are not controlling for this endogenous matching. In
contrast, when we use xed eects we can control for the fact that bank-rm
matching is not random by comparing 1banks and 1banks behavior
keeping constant and homogenous the level of default risk between banks
types. It is only in this last situation that we are able to compare 1banks
with 1banks perfectly, since then the interest rate spread between these
two types of banks (and their dierent lending behavior) only reects their
dierent roles in the credit market. In sum, the interest spread between
1banks and 1banks in good times reported in Figure 4 (using xed ef-
fects) is an unbiased measure of the premium that rms pay in order to get
a loan restructuring in bad times.
27
6 Comparing dierent theories of relation-
ship banking
The relationship banking literature distinguishes between dierent benets
from a long-term banking relation. Except for Berlin and Mester (1999)
the other theories do not explicitly consider how relationship lending would
evolve over the business cycle. Nevertheless it is instructive to briey compare
our ndings with the likely predictions of the other main theories.
We have identied four dierent strands of relationship banking theories,
which dier in their predictions of default rates, cost of credit, and credit
availability over the business cycle. A rst strand emphasizes the role of
1Banks in providing (implicit) insurance to rms towards future access to
credit and future credit terms (Berger and Udell, 1992; Berlin and Mester
1999). According to this (implicit) insurance theory, 1banks do not have
better knowledge about rms types and therefore they should experience
similar default rates (in crisis times) to those of 1banks. Although Berlin
and Mester nd that banks funded more heavily with core deposits provide
more loan-rate smoothing in response to exogenous changes in aggregate, it
does not follow from this nding that the rms with the lowest credit risk
choose this loan-rate smoothing service.
A second strand underscores the monitoring role of 1banks (Holmstrom
and Tirole 1997, Boot and Thakor 2000, Hauswald and Marquez 2000). Ac-
cording to the monitoring theory, only rms with low equity capital choose
a monitored bank loan from an 1bank, while rms with sucient cash
(or collateral) choose cheaper loans from a 1bank. By this theory adverse
selection is a minor issue, and monitoring is simply a way to limit the rms
interim moral hazard problems. Although these monitoring theories do not
make any explict predictions on default rates in a crisis, it seems plausible
that these theories would predict higher probabilities of default in bad times
for rms borrowing from 1banks.
A third strand plays up the greater ex-ante screening abilities (of new
loan applications) of 1-banks due to their access to both hard and soft in-
formation about the rm (Agarwal and Hauswald 2010, Puri et al. 2010).
A plausible prediction of these ex-ante screening theories would seem to be
that 1banks have lower default rates in crisis times than 1banks. An-
other plausible prediction is that 1banks would benet from an ex-post
monopoly of information both in good and bad times, thus charging higher
28
lending rates than 1banks in both states on the loans they roll over.
A fourth strand of relationship banking theories, on which our model
builds, emphasizes (soft) information acquisition about borrowers types over
time. This role is closer to the one emphasized in the original contributions by
Sharpe (1990), Rajan(1992) and Von Thadden (1995). This strand of theories
puts the 1bank in the position of oering continuation lending terms that
are better adapted to the specic circumstances in which the rm may nd
itself in the future. This line of theories predicts that 1banks charge higher
lending rates in good times on the loans they roll over, and lower rates in
bad times to help their best clients through the crisis. In contrast, 1banks
oer cheaper loans in good times but roll over fewer loans in bad times. Also,
according to this theory we should observe lower delinquency rates in bad
times for 1banks that roll over their loans.
As a rst step in the comparison among these dierent theoretical models,
we focus on the relationship between the probability for a rm / to go into
default and the composition of its transactional vs relationship nancing.
From our test of Proposition 1 (see equation 11) we nd that 1banks
exhibit higher default rates in bad times. This nding is consistent with the
predictions of our theory as well the third strand of theories based on ex-ante
screening.
The comparison of interest rates of 1banks and 1banks in good times
and in bad times provides additional insights into the likely benets of re-
lationship banking. In particular, our ndings that 1banks charge higher
rates than 1banks in good times and vice-versa in bad times are only con-
sistent with the prediction of our theory. Importantly, we do not observe an
ex-post monopoly of information for 1banks such that 1banks always
charge higher rates than 1banks on the loans they roll over.
7 Conclusion
The theoretical approach we suggest, which emphasizes the idea that rela-
tionship lending allows banks to learn rms types and therefore help them
in bad types, provides predictions on relative rates of default in a crisis,
the behavior of interest rates, and the amount of equity capital of relation-
ship banks that the our data broadly supports. Our empirical analysis thus
provides further guidance on what relationship-lending achieves in the real
economy. We have found that relationship banking is an important mitigat-
29
ing factor of crises. By helping protable rms to retain access to credit in
times of crisis relationship banks dampen the eects of a credit crunch. How-
ever, the role relationship banks can play in a crisis is limited by the amount
of excess equity capital they are able to hold in anticipation of a crisis. Banks
entering the crisis with a larger equity capital cushion are able to perform
their relationship banking role more eectively. These results are consistent
with other empirical ndings for Italy (see, amongst others, Albertazzi and
Marchetti (2010) and Gobbi and Sette (2012)).
Our analysis suggests that if more rms could be induced to seek a long-
term banking relation, and if relationship banks could be induced to hold a
bigger equity capital buer in anticipation of a crisis, the eects of crises on
corporate investment and economic activity would be smaller. However, ag-
gressive competition by less well capitalized and lower-cost transaction banks
is undermining access to relationship banking. As these banks compete more
aggressively more rms will switch away from 1banks and take a chance
that they will not be exposed to a crisis. And the more rms switch the
higher the costs of 1banks. Overall, the ercer competition by 1banks
contributes to magnifying the amplitude of the business cycle and the pro-
cyclical eects of bank capital regulations.
References
[1] Agarwal, S. and Hauswald R. (2010), Distance and Private Information
in Lending, Review of Financial Studies, 23, 2757-2788.
[2] Albertazzi, U. and Marchetti D. (2010), Credit Supply, Flight to Qual-
ity and Evergreening: An Analysis of Bank-Firm Relationships after
Lehman, Bank of Italy, Temi di discussione, 756.
[3] Angelini, P., Di Salvo P. and Ferri G. (1998), Availability and Cost of
Credit for Small Businesses: Customer Relationships and Credit Coop-
eratives, Journal of Banking and Finance, 22, 925-954.
[4] Berger, A.N., Miller N.H., Petersen M.A. , Rajan R.G., Stein J.C.(2005),
Does function follow organizational form? Evidence from the lending
practices of large and small banks, Journal of Financial Economics,
76(2), 237-269.
30
[5] Berger, A.N. and Udell G.F. (1992), Some Evidence on the Empirical
Signicance of Credit Rationing, Journal of Political Economy, 100(5),
1047-1077.
[6] Berger, A.N. and Udell G.F. (1995). Relationship Lending and Lines
of Credit in Small Firm Finance, Journal of Business, 68, 351379.
[7] Berger, A.N. and Udell, G.F. (2004), The Institutional Memory Hy-
pothesis and the Procyclicality of Bank Lending Behavior, Journal of
Financial Intermediation, 13, 458-495.
[8] Berger, A.N. and Udell, G.F. (2006), A More Complete Conceptual
Framework for SME Finance, Journal of Banking and Finance, 30,
29452966.
[9] Berlin, M. and Mester, L. (1999), Deposits and Relationship Lending,
Review of Financial Studies, 12(3), 579-607.
[10] Bolton, P. and Scharfstein, D.S. (1996), Optimal Debt Structure and
the Number of Creditors, Journal of Political Economy, 104, 125.
[11] Bolton, P. and X. Freixas (2000), Equity, Bonds and Bank Debt: Cap-
ital Structure and Financial Market Equilibrium under Asymmetric In-
formation, Journal of Political Economy, 108, 324351.
[12] Bolton, P. and Freixas X. (2006), Corporate Finance and the Monetary
Transmission Mechanism, Review of Financial Studies, 3, 829870.
[13] Bonaccorsi di Patti, E. and Sette, E. (2012), Bank Balance Sheets and
the Transmission of Financial Shocks to Borrowers: Evidence from the
2007-2008 Crisis, Bank of Italy, Temi di discussione, 848.
[14] Boot, A.W.A. (2000), Relationship banking: What do we know?,
Journal of Financial Intermediation, 9, 7-25.
[15] Boot, A.W.A. and A. Thakor, (2000), Can Relationship Banking Sur-
vive Competition?, Journal of Finance, American Finance Association,
55(2), 679-713.
[16] Cremer, J., Garicano L., and Prat A. (2007), Language and the Theory
of the Firm, Quarterly Journal of Economics, 1, 373-407.
31
[17] De Mitri, S., Gobbi G. and Sette E. (2010), Relationship Lending in a
Financial Turmoil, Bank of Italy, Temi di discussione, 772.
[18] Degryse, H. and Ongena S. (2005), Distance, Lending Relationships,
and Competition, Journal of Finance, 60(1), 231-266.
[19] Detragiache, E., Garella, P. and Guiso, L. (2000), Multiple versus Sin-
gle Banking Relationships: Theory and Evidence, Journal of Finance,
55, 1133-1161.
[20] Diamond, Douglas W., and Raghuram G. Rajan (2000), A theory of
bank capital, Journal of Finance, 55, 24312465.
[21] Diamond, Doulas. W. and Raghuram Rajan (2001), Liquidity Risk,
Liquidity Creation, and Financial Fragility: A Theory of Banking,
Journal of Political Economy, 109: 289-327.
[22] Diamond, Douglas W., and Raghuram Rajan (2005), Liquidity Short-
age and Banking Crises, Journal of Finance, 60: 615647.
[23] Farinha, L.A. and Santos J.A.C. (2002), Switching from Single to
Multiple Bank Lending Relationships: Determinants and Implications,
Journal of Financial Intermediation, 11(2), 124151.
[24] Foglia, A., Laviola, S. and Marullo Reedtz, P. (1998), Multiple Bank-
ing Relationships and the Fragility of Corporate Borrowers, Journal of
Banking and Finance, 22, 1441-1456.
[25] Gambacorta, L. and Mistrulli P. (2004), Does Bank Capital Aect
Lending Behavior?, Journal of Financial Intermediation, 13, 436457.
[26] Gambacorta, L. and Mistrulli P. (2013), Bank Heterogeneity and Inter-
est Rate Setting: What Lessons have we Learned since Lehman Broth-
ers?, Journal of Money Credit and Banking, forthcoming.
[27] Gobbi, G. and Sette E. (2012), Relationship Lending in a Financial
Turmoil, Bank of Italy, mimeo.
[28] Hale, G. and Santos, J.A.C. (2009), Do banks price their informational
monopoly?, Journal of Financial Economics, 93, 185-206.
32
[29] Hauswald, R. and R. Marquez (2000), Relationship Banking, Loan
Specialization and Competition, Proceedings, Federal Reserve Bank of
Chicago, May, 108-131.
[30] Holmstrom, B. and Tirole J. (1997), Financial Intermediation, Loan-
able Funds and the Real Sector, Quarterly Journal of Economics,
112(3), 663691.
[31] Houston, J. and James C. (1996), Bank Information Monopolies and
the Mix of Private and Public Debt Claims, The Journal of Finance,
51(5), 18631889.
[32] Kashyap, A. and Stein J.C. (1995), The Impact of Monetary Policy on
Bank Balance Sheets, Carnegie Rochester Conference Series on Public
Policy, 42, 151-195.
[33] Kashyap, A. and Stein J.C. (2000), What Do a Million Observations
on Banks Say About the Transmission of Monetary Policy, American
Economic Review, 90, 407-428.
[34] Kishan, R.P. and Opiela T.P. (2000), Bank Size, Bank Capital and the
Bank Lending Channel, Journal of Money, Credit and Banking, 32,
121141.
[35] Panetta, F., Angelini P. (coordinators), Albertazzi, U., Columba, F.,
Cornacchia, W., Di Cesare, A., Pilati, A., Salleo, C., Santini, G., (2009),
Financial sector pro-cyclicality: lessons from the crisis, Bank of Italy
Occasional Papers Series, 44.
[36] Petersen, M.A. (2009), Estimating Standard Errors in Finance Panel
Data Sets: Comparing Approaches, Review of Financial Studies, 22,
435480.
[37] Petersen, M. and Rajan R. (1994), The Benets of Lending Relation-
ships: Evidence from Small Business Data, The Journal of Finance,
49, 337.
[38] Puri, M., Rocholl J. and Sascha S. (2010) On the Impor-
tance of Prior Relationships in Bank Loans to Retail Cus-
tomers; mimeo, SSRN: http://ssrn.com/abstract=1572673 or
http://dx.doi.org/10.2139/ssrn.1572673
33
[39] Rajan, R. (1992), Insiders and Outsiders: The Choice Between In-
formed Investors and Arms Lenght Debt, Journal of Finance, 47(4),
13671400.
[40] Schularick, M. and Taylor A.M. (2011), Credit booms gone bust: mon-
etary policy, leverage and nancial crises, 1870-2008, American Eco-
nomic Review, forthcoming.
[41] Sharpe, S.A. (1990), Asymmetric Information, Bank Lending and Im-
plicit Contracts: A Stylized Model of Customer Relationships, Journal
of Finance, 45(4), 10691087.
[42] Stein, J.C. (2002), Information Production and Capital Allocation: De-
centralized Versus Hierarchical Firms, Journal of Finance, 57, 1891-
1921.
[43] Von Thadden, E. L. (1995), Long-term contracts, short-term invest-
ment and monitoring, Review of Economic Studies, 62, 557575.
34
Appendix A. Mathematical proofs
Proof of Proposition 1
We shall characterize the equilibrium lending terms and loan renancing
using backwards induction. These lending terms and roll-over decisions will
depend on whether we are considering a safe rm for which condition (3)
holds (j b j) or a risky rm (j < b j).
If the project is successful, rms are able to repay their loan out of
their cash ow \
1
. This occurs with probability j
S
. In this case the
rm continues to period 2 and gets \
1
if it is an Hrm.
If the project fails at time t = 1, rms with j b j will be able to roll
over their debts. Their debt will then be rolled over against a promised
repayment of :
S
T
(j
S
) that reects state of nature o. When with j b j,
Hrms are able to make suciently high promised expected repay-
ments i
1
:
1
T
(j
1
) even in the recession state, so that for these rms we
have :
T
(j) given by the break-even condition:
:
T
(j) = 1 +j
T
.
If, instead j < b j, liquidation occurs in state 1 if the rm is not suc-
cessful, which happens with probability o(1 j
1
). The gross interest
rate :
T
(j) is then given by the break even condition:
[j + (1 o)(1 j
G
)] :
T
(j) +o(1 j
1
)\
1
= 1 +j
T
.
Proof of Proposition 2
Hrms are then able to secure new lending at gross interest rate :
1
1
=
,\
1
in both recession and boom states. Under these conditions, the rst
period gross interest rate :
1
(j) is given by the break-even condition:
j:
1
(j) + (1 j)[i(1 :),\
1
+ (1 i)\
1
] = 1 +j
1
.
where
16
i
(1 o)(1 j
G
)i
G
+o(1 j
1
)i
1
(1 j)
16
We assume again that the intermediation cost of dealing with an 1bank is entirely
capitalized in period 0.
35
Proof of Proposition 3
If 1banks have no incentive to roll over the rms joint debts of Hrms,
then the benets of combining the two types of debt are lost and the rm
would be better o with 100% 1bank nancing. Consequently, the combi-
nations of the two types of debt, 1
1
and 1
T
. is of interest only in so far as
the 1bank has an incentive to use its information to restructure the debts
of unsuccessful Hrms.
This means that combining both types of debts only makes sense if the
following constraint is satised:
,\
1
(1 :) :
1T
T
1
T
1
1
\
1
. (19)
The LHS represents what the 1bank obtains by rolling over all the period
t = 1 debts of an unsuccessful Hrm. When there is a combination of
1debt and 1debt, a roll-over requires not only that the 1bank extends
a new loan to allow the rm to repay :
1T
1
at t = 1, but also that it extends a
loan to allow the rm to repay :
1T
T
to the 1bank. As a result, the 1bank
can hope to get only ,\
1
(1 :) :
1T
T
1
T
by rolling over an unsuccessful
Hrms debts. This amount must be greater than what the 1bank can
get by liquidating the rm at t = 1. namely 1
1
\
1
.
1banks know that if they are lending to an Hrm their claim will
be paid back by the 1bank, provided the above condition (19) is met. So
they will obtain the par value, :
1T
T
for sure if they lend to an Hrm and a
fraction 1
T
of the residual value \
1
if, instead the rm is an 1rm. The
corresponding rate is therefore:
:
1T
T
=
(1 +j
T
) (1 j)(1 i)\
1
j + (1 j)i
. (20)
As intuition suggests, constraint (19) holds only if the amount of 1bank
debt the rm takes on is below some threshold. To establish this, note that
replacing 1
1
= 1 1
T
condition (19) can be rewritten as:
,(1 :)\
1
\
1
1
T
(:
1T
T
\
1
). (21)
Substituting for :
1T
T
we obtain that the following maximum amount of
transaction lending is consistent with ecient restructuring:
1
T

(1 +j
T
) (1 j)(1 i)\
1
j + (1 j)i
\
1

,\
1
(1 :) \
1
. (22)
36
which simplies to:
1
T

(1 +j
T
) \
1
j + (1 j)i

,\
1
(1 :) \
1
(23)
Implying that:
1
T

(j + (1 j)i)

,\
1
(1 :) \
1

1 +j
T
\
1
.
As the rm optimally chooses the amounts 1
T
and 1
1
. it will choose the
combination that maximizes
1T
, which is equivalent to minimizing the total
funding cost c
c = j :
1T
1
(j)(1 1
T
) +j :
1T
T
1
T
under the constraint (19) that guarantees that the 1bank has an incentive
to restructure Hrms.
The expression for c can be simplied by using the break even constraint
for the 1bank, which is given by:
j :
1T
1
(1 1
T
) + (24)
(1 j)

i((1 :),\
1
:
1T
T
1
T
) + (1 i)(1 1
T
)\
1

= (1 +j
1
)(1 1
T
)
or,
j :
1T
1
(1 1
T
) =
(1 +j
1
)(1 1
T
) (1 j)

i((1 :),\
1
:
1T
T
1
T
) + (1 i)(1 1
T
)\
1

Collecting terms in 1
T
on the right hand side we then get:
j :
1T
1
(1 1
T
) =
1
T
[(1 +j
1
) + (1 j)(i:
1T
T
+ (1 i)\
1
)] + (1 +j
1
) (25)
(1 j)

i(1 :),\
1
+ (1 i)\
1

Replacing j:
1T
1
(11
T
) by its value in (25) and ignoring constant factors
we thus obtain the equivalent funding cost minimization problem :
min
1

[(1 +j
1
) + ((1 j)i +j):
1T
T
+ (1 j)(1 i)\
1
)]1
T
37
But notice that the coecient
[(1 +j
1
) + ((1 j)i +j):
1T
T
+ (1 j)(1 i)\
1
)] 0
as :
1T
T
satises
((1 j)i +j):
1T
T
+ (1 j)(1 i)\
1
= 1 +j
T
and j
1
j
T
.
Consequently the condition (19) is always binding. This allows to replace
1
T
in (24) leading to:
j:
1T
1
(1 1
T
) + (1 j)\
1
= (1 +j
1
)(1 1
T
) (26)
thus obtaining the expression for :
1T
1
.
Proof of Proposition 4
Let =
T

1T
denote the dierence in expected payos for an
Hrm from choosing 100% 1nancing over mixed nancing, where

T
= j(2\
1
:
T
(j)) +(1o)(1j
G
)(\
1

:
T
(j)
i
G
) +o(1j
1
)(\
1

:
T
(j)
i
1
)
for j b j, where :
T
(j) = 1 +j
T
and

T
= j(2\
1
:
T
(j)) + (1 o)(1 j
G
)(\
1

:
T
(j)
i
G
)
for j < b j, where
:
T
(j) =
1 +j
T
o(1 j
1
)\
1
oj
1
+ 1 o
and,

1T
= j(2\
1
:
1T
1
(j)(1 1
T
) :
1T
T
(j)1
T
) + (1 j)(1 ,)\
1
Consider rst the case j b j
Combining these expressions can be written as follows:
(j) = j(:
1T
1
(j)(1 1
T
) +:
1T
T
(j)1
T
:
T
(j)) (27)
+(1 o)(1 j
G
)[,\
1

:
T
(j)
i
G
] +
+o(1 j
1
)(,\
1

:
T
(j)
i
1
)
38
The rst term,
j(:
1T
1
(1 1
T
) +:
1T
T
(j)1
T
) :
T
(j)).
reects the dierence in the costs of funding when the rm is successful,
which occurs with probability j. The other terms measure the dierence for
a non successful rm between the benets of relationship banking and those
of transactional banking.
To simplify the expression for (j) let
j[:
1T
1
(j)(1 1
T
) +:
1T
T
(j)1
T
]
From the break even condition (24) we then obtain that
= (1 +j
1
)(1 1
T
) + (1 j)i :
1T
T
(j)1
T
+j :
1T
T
(j)1
T
(1 j)

i(1 :),\
1
+ (1 i)(1 1
T
)\
1

Substituting for
:
1T
T
=
(1 +j
T
) (1 j)(1 i)\
1
j + (1 j)i
the above expression simplies to:
= (1+j
1
)(j
1
j
T
)1
T
(1j)

i(1 :),\
1
+ (1 i)(1 1
T
)\
1

(1j)(1i)1
T
\
1
(28)
Substituting for in (j) we obtain:
(j) = j :
T
(j) + (1 o)(1 j
G
)[(1 ,)\
1

:
T
(j)
i
G
] +
o(1 j
1
)

,)\
1

:
T
(j)
i
1

we obtain:
39
(j) = (1 +j
1
) (j
1
j
T
)1

T
(29)
(1 j)

i(1 :),\
1
+ (1 i)\
1

j(1 +j
T
) + (1 j),\
1
(1 +j
T
)[
(1 o)(1 j
G
)
i
G
+
o(1 j
1
)
i
1
]
Dierentiating with respect to j
1
and noting that
dj
G
dj
1
=
dj
dj
1
= 1
and that:
d1

T
dj
=
(1 i)

,\
1
(1 :) \
1

1 +j
T
d(j)
dj
1
= (j
1
j
T
)
d1

T
dj
1

d(1 j)i
dj
1

(1 :),\
1
\
1

(30)
+\
1
(1 +j
T
) ,
A
\
1
+(1 +j
T
)[
(1 o)
i
G
+
o
i
1
]
Using
d(1 j)i
dj
1
= [(1 o)i
G
+oi
1
]
and
d1

T
dj
1
=

(1 :),\
1
\
1

1 +j
T
(1 [(1 o)i
G
+oi
1
])
we further obtain:
40
d(j)
dj
1
= (j
1
j
T
)

(1 :),\
1
\
1

1 +j
T
(1 [(1 o)i
G
+oi
1
])(31)
+[(1 o)i
G
+oi
1
]

(1 :),\
1
\
1

+\
1
(1 +j
T
) ,\
1
+(1 +j
T
)[
(1 o)
i
G
+
o
i
1
]
Or, equivalently,
d(j)
dj
1
= (j
1
j
T
)

(1 :),\
1
\
1

1 +j
T
(1 [(1 o)i
G
+oi
1
])(32)
[(1 o)i
G
+oi
1
] :,\
1
(,\
1
\
1
)(1 [(1 o)i
G
+oi
1
])
+(1 +j
T
)[
(1 o)
i
G
+
o
i
1
1]
Now, under assumption A1 the rst two terms are negligeable, while
under assumption A2 the last two terms are positive, leading to
o(j)
oj

0 .
Next, consider the case j < b j.
Proceeding as before, can be written as follows:
(j) = j(:
1T
1
(j)(1 1
T
) +:
1T
T
(j)1
T
:
T
(j)) (33)
+(1 o)(1 j
G
)[,\
1

:
T
(j)
i
G
] +
o(1 j
1
)(1 ,)\
1
We will simply show that A1 and A2 are sucient conditions for (j) <
0
41
The rst term,
j(:
1
(j)(1 1
T
) +:
1T
T
(j)1
T
) :
T
(j)).
reects the dierence in the costs of repaying the loan when the rm is suc-
cessful, which occurs with probability j. The other terms measure the dif-
ference for a non successful rm between the benets of relationship banking
and those of transactional banking.
To simplify the expression for (j) let
j[:
1T
1
(j)(1 1
T
) +:
1T
T
(j)1
T
]
From the break even condition (24) we then obtain that
= (1 +j
1
)(1 1
T
) + (1 j)i:
1T
T
(j)1
T
+j :
1T
T
(j)1
T
(1 j)

i(1 :),\
1
+ (1 i)(1 1
T
)\
1

Substituting for
:
1T
T
=
(1 +j
T
) (1 j)(1 i)\
1
j + (1 j)i
the above expression simplies to:
= (1 +j
1
) (j
1
j
T
)1
T
(1 j)

i(1 :),\
1
+ (1 i)\
1

(34)
Substituting for in (j) we obtain:
(j) = j :
T
(j) + (1 o)(1 j
G
)[(,\
1

:
T
(j)
i
G
] +
o(1 j
1
)

(1 ,)\
1

As i
G
< 1. the expression j :
T
(j)+(1o)(1j
G
)
v

(j)
i

has a lower bound


= :
T
(j) [j + (1 o)(1 j
G
)]
but j + (1 o)(1 j
G
) = oj
1
+ 1 o. so that replacing :
T
(j) we obtain
= 1 +j
T
o(1 j
1
)\
1
42
As a consequence, we obtain
(j) < + (1 o)(1 j
G
),\
1
+
o(1 j
1
)

(1 ,)\
1

which after replacement of and leads to


(j) < (j
1
j
T
)(1 1

T
) (35)
(1 j)

i(1 :),\
1
+ (1 i)\
1

+ (1 j),\
1
+o(1 j
1
)\
1
o(1 j
1
)\
1
Rearranging terms this expression becomes:
(j) < (j
1
j
T
)(1 1

T
) + (1 j):,\
1
(36)
(1 j)

i,\
1
+ (1 i)\
1

+ (1 j),\
1
o(1 j
1
)(\
1
\
1
)
that is
(j) < (j
1
j
T
)(1 1

T
) + (1 j):,\
1
(37)
+(1 j)(1 i)

,\
1
\
1

o(1 j
1
)(\
1
\
1
)
Under A1 the rst two terms are small. Under A2

,\
1
\
1

is also
small so that the last term dominates and (j) < 0.
43
Appendix B. Technical details regarding the
data
We construct the database by matching four dierent sources.
i) The Credit Register (CR) containing detailed information on all
loan contracts granted to each borrower (i.e. the amount lent, the type of
loan contract, the tax code of the borrower).
ii) The Bank of Italy Loan Interest Rate Survey, including information
on interest rates charged on each loan reported to the CR and granted by
a sample of more than 200 Italian banks; this sample accounts for more
than 80% of loans to non-nancial rms and is highly representative of the
universe of Italian banks in terms of bank size, category and location. We
investigate overdraft facilities (credit lines) for three main reasons. First,
this kind of lending represents the main liquidity management tool for rms
especially the small ones (with fewer than 20 employees) that are prevalent
in Italy which cannot aord more sophisticated instruments. Second, since
these loans are highly standardized among banks, comparing the cost of
credit among rms is not aected by unobservable (to the econometrician)
loan-contract-specic covenants. Third, overdraft facilities are loans granted
neither for some specic purpose, as is the case for mortgages, nor on the
basis of a specic transaction, as is the case for advances against trade credit
receivables (Berger and Udell, 1995).
iii) The Supervisory Reports of the Bank of Italy, from which we ob-
tain the bank-specic characteristics (size, liquidity, capitalization, funding
structure). Importantly, for all the banks in the sample, we obtain informa-
tion on the credit concentration of the local credit market in June 2008. We
compute Herndahl indexes for each province (similar to counties in the US)
using the data on loans granted by banks.
iv) The CERVED database, which includes balance sheet information
on about 500,000 companies, mostly privately owned. Balance sheet data are
taken at t 1. This is important since credit decisions in t on how to set
rms interest rates on credit lines are based on balance sheet information
that has typically a lag.
17
We match these four sources obtaining a dataset of bank-rm lending re-
17
For more information, see http://www.cerved.com/xportal/web/eng/aboutCerved/aboutCerved.jsp.
The methodology for the calculation of the 7score, computed annually by CERVED, is
provided in Altman et al. (1994).
44
Table B1 Summary statistics
Zscorein2008:Q4 Obs.
Tbank
(1)
Credit
History
(2)
LTD
Log
Loans
Spread
2007:Q2
(3)
Spread
2010:Q1
(3)
1=Safe 4,045 0.68 10.92 0.991 7.48 3.81 5.38
2=Solvent 7,968 0.69 10.36 0.995 7.65 3.94 5.65
3=Vulnerable 67,614 0.71 10.33 0.981 7.89 4.39 6.33
4=Risky 106,697 0.72 9.35 0.963 7.91 4.88 7.33
Total 186,324 0.72 9.78 0.971 7.88 4.64 6.86
Note:(1)Shareofloansthatisgrantedbyabankthathasitsheadquarteroutsidethesameprovincewhere
thefirmhasitsheadquarter.(2)Numberofyearselapsedsincethefirsttimeaborrowerwasreportedto
theCreditregister.(2)Interestrateoncreditlinesminusonemonthinterbankrate.
lationships. In the paper we focus on multiple lending by selecting those rms
which have a credit line with at least two Italian banks in June 2008. This
limits the analysis to 216,000 observations. However, around 80% of Italian
non-nancial rms have multiple lending relationships, so this selection does
not limit our study from a macroeconomic point of view.
We clean outliers from the data, cutting the top and bottom fth per-
centile of the distribution of the dependent variables we use in the regression.
An observation has been dened as an outlier if it lies within the top or bot-
tom fth percentile of the distribution of the dependent variables (:
),I
and
1
),I
). After these steps our sample reduces to around 185,000 observations
(75,000 rms), which we use for the empirical analysis. The following table
gives some basic information on the main variables used in the regressions.
Appendix C. Robustness checks
We have checked the robustness of the results in several ways.
(1) Main bank. As there is not a clear consensus on the way rela-
tionship characteristics are identied, we have tested the robustness of the
results by including in the baseline regressions an additional measure of re-
lationship banking, namely a dummy for the main bank. This dummy
typically captures incentive to monitor eects or skin in the game ef-
fects. In particular, we have rst calculated the share of loans granted by
45
each bank to the rm and constructed two variables: i) the highest share
of lending granted by the main bank (`cr:/); ii) a dummy (`ci:) that
is equal to one if that bank grants the highest share of lending to the rm.
However, as in several cases many banks had a pretty low and similar share
of total lending, we have decided to consider as main bank only those -
nancial intermediaries that granted not only the highest share but also at
least one quarter of the total loans.
We have therefore modied equations (11)-(13) for the marginal probit
model, the interest rate (:
),I
) and outstanding loans in real terms (1
),I
) in
the following way:
`1(Firm ks default= 1) = c++:1:/c:c
I
+``cr:/
I
+i(1:/c:c
I
`cr:/
I
)+
I
(11)
:
),I
= i +,+1 /c:/
),I
+`ci:
),I
+t(1 /c:/
),I
`ci:
),I
)+
),I
(12)
1
),I
= o +c+j1 /c:/
),I
+t`ci:
),I
+o(1 /c:/
),I
`ci:
),I
)+
),I
(13)
where c, i and o are bank-xed eects, is a vector of industry xed eects,
, and c are rm-xed eects.
The results reported in Table C1.1 indicates that the highest the share
of loan granted to a rm the lower is the probability that the rm goes
into default. At the same time, the eect of transactional loans on default
probability is still in place and similar in magnitude with respect to that in
Table 3. In particular, the probability for a rm to go into default increases
with the share of 1bank nancing and reach the maximum of around 0.4%
when 1 :/c:c
I
is equal to 1. This result remains pretty stable by enriching
the set of controls with additional rm-specic characteristics (see panel II in
Table C1) or by calculating the proportion of transactional loans 1 :/c:c
I
not in value but in terms of the number of banks that nance rm k (see
panel III in Table C1).
The main results of our work remain also with respect to the two cross-
sectional equations (12) and (13). In line with the predictions of the model,
Table C2 indicates that 1banks (compared to 1banks) provide loans at
a cheaper rate in good time and at a higher rate in bad time (see columns
46
I and II). As for loan quantities, other things being equal, T-banks always
provide on average a lower amount of lending, especially in bad times (see
columns III and IV). Interestingly, we nd that a bank with a high share
of lending to a given rm tends to grant always lower interest rates and to
further reduce the cost of credit by more in time of crisis. However, we also
nd that the main bank reduce the amount of loans in a crisis. This nding
is consistent with the result in Gambacorta and Mistrulli (2013) and may
be interpreted as the eect of more bank risk aversion and a greater need to
diversify credit risk following the crisis.
(2) Region instead than province. One possible objection to the denition
used for the relationship dummy 1bank is that considering the bank and
the rm as "close" only if both have headquarter in the same province could
be too restrictive. For example, banks may be able to get soft information,
i.e. information that is dicult to codify, which is a crucial aspect of lending
relationships, also if they are headquartered inside the same region where the
rm has is main seat.
We have therefore replicated the results of Table 3 and Table 4 in the main
text by using a dierent denition for relationship and transaction banks. In
particular, the 1bank dummy is equal to 1 if rm / is headquartered in the
same region (instead than the province) where bank , has its headquarters;
1bank is equal to 1 if 1bank=0. Results reported in Tables C3 and C4
are very similar to those in the main text. Interestingly, the absolute values
of coecients are slightly reduced pointing to the fact that informational
asymmetries increase with functional distance.
(3) All foreign banks are T-banks. In the paper we divide 1banks and
1banks according to the distance between the lending bank headquarters
(at the single bank level, not at the group level) and rm headquarters.
This raises some questions for foreign banks (subsidiaries and branches of
foreign banks). Following this denition, branches of foreign banks are always
1banks. This classication is correct because lending strategic decisions are
typically taken by the banks headquarter located outside Italy.
However, these loans have not a big weigh in the database and represent
only 0.04% of the cases. On the contrary, subsidiaries of foreign banks are
treated as the Italian banks. This hypothesis seems plausible as these banks
have legal autonomy and are subject to Italian regulation. However, to test
the robustness of the results we have therefore replicated the estimations
reported in Table 3 and Table 4 by imposing that all foreign bank head-
quartered in Italy and with legal autonomy (around 7% of observations) are
47
1banks. This means that the 1bank dummy is equal to 1 if rm / is
not headquartered in the same province where bank , has its headquarters
or bank , is a foreign bank; 1bank is equal to 1 if 1bank= 0. Even in
this robustness test results are very similar to the baseline case (see Tables
C5 and C6).
(4) New rms. One of the main hypothesis of the model is that at t = 0
no bank can distinguish rms type. To make the empirical part closer to the
theoretical onet we have therefore estimated equations (11)-(13) on a subset
of around 6,000 new rms, that entered the credit register in the period
2005:Q2:2007:Q2. The results are qualitatively very similar to that obtained
from the baseline equations (see Tables C7 and C8).
48
49

Figure 1
Average firm cash flows in state S

Figure 2
100% Transactional Banks Payoff
50


Figure 3
Bank lending, interest rates and the business cycle in Italy
(a) Bank lending to the private sector
1, 2

5
0
5
10
15
20
2003 2004 2005 2006 2007 2008 2009 2010
Medium and large non-financial firms
Small non-financial firms
Households

(b) Interest rate on overdraft and interbank rate
1, 3

0
2
4
6
8
2003 2004 2005 2006 2007 2008 2009 2010
Interest on overdrafts (a)
3-month interbank rate (b)
Spread (a) - (b)

(c) Real GDP and stock market capitalization
4, 5

750
1,000
1,250
1,500
1,750
2,000
2,250
345
350
355
360
365
370
375
2003 2004 2005 2006 2007 2008 2009 2010
Real GDP (right axis)
Stock market capitalization

Notes. The vertical line indicates Lehmans default.
1
Monthly data.
2
Annual growth rates. Bad loans are
excluded. The series are corrected for the impact of securitization activity.
3
Percentage points. Current
account overdrafts are expressed in euro.
4
Quarterly data.
5
Real GDP in billions of euro. Stock market
capitalization refers to the COMIT Globale Index, 31 Dec. 1972 = 100.
Sources: Bank of Italy; Bloomberg.
51


Figure 4
Lending supply and interest rate setting by banks type and state of the world
1

(a1) Interest rate: good times (2007:q2) (a2) Interest rate: bad times (2010:q1)
(b1) Lending: good times (2007:q2) (b2) Lending: bad times (2010:q1)
1
This figure reports a graphical representation of the results in Table 5. The horizontal axis of each graph reports the Z-score,
an indicator of the probability of default of firms. These scores can be mapped into four levels of risk: 1) safe; 2) solvent; 3)
vulnerable; 4) risky. The vertical axis of graphs (a1) and (a2) indicate the level of the interest rate applied by the two bank types
on credit lines to the 4 different kinds of firms; those of graphs (b1) and (b2) report the log of lending in real terms supplied by
the two bank types.

52

Figure 5
Graphical analysis of the results in Table 5 without fixed effects
1

(a1) Interest rate: good times (2007:q2) (a2) Interest rate: bad times (2010:q1)
(b1) Lending: good times (2007:q2) (b2) Lending: bad times (2010:q1)
1
This figure reports a graphical representation of the results obtained re-running the same models reported in Table 5 without
fixed effects. The horizontal axis of each graph reports the Z-score, an indicator of the probability of default of firms. These
scores can be mapped into four levels of risk: 1) safe; 2) solvent; 3) vulnerable; 4) risky. The vertical axis of graphs (a1) and
(a2) indicate the level of the interest rate applied by the two bank types on credit lines to the 4 different kinds of firms; those of
graphs (b1) and (b2) report the log of lending in real terms supplied by the two bank types.
53

Table 1 Descriptive statistics. Bank-firm relationship
Bank-firm loan types


Obs.


%


Spread
good time
(2007:q2)
(a)
Spread
bad time
(2010:q1)
(b)

(b) -(a)
Log Loans
good time
(2007:q2)
(c )
Log Loans
bad time
(2010:q1)
(d)

(d) -(c)
Capital to
asset ratio
(2007:q2)
(e)
Capital to
asset ratio
(2010:q1)
(f)
(f)-(e)


ALL FIRMS

i) Relationship only 18693 10.1% 4.3 6.2 1.9 7.74 7.73 -0.011 9.103 8.794 -0.31
ii) Both types 84598 45.8% 4.5 6.7 2.2 7.96 8.00 0.036 8.843 8.743 -0.10
iii) Transactional only 81604 44.1% 4.8 7.1 2.3 7.78 7.81 0.029 8.547 8.793 0.25
Total 184895 100.0% 4.6 6.8 2.2 7.86 7.89 0.028 8.739 8.770 0.03
H-FIRMS
i) Relationship only 18489 10.1% 4.2 6.2 1.9 7.74 7.73 -0.006 9.096 8.79 -0.30
ii) Both types 84129 45.9% 4.5 6.7 2.2 7.95 7.99 0.039 8.543 8.56 0.02
iii) Transactional only 80493 44.0% 4.8 7.1 2.3 7.77 7.80 0.032 8.842 8.79 -0.05
Total 183111 100.0% 4.6 6.8 2.2 7.85 7.88 0.031 8.730 8.69 -0.04
L-FIRMS
i) Relationship only 206 11.6% 6.0 9.0 3.0 8.07 7.90 -0.169 8.98 8.70 -0.28
ii) Both types 439 24.6% 5.9 9.4 3.5 8.54 8.33 -0.207 8.949 9.04 0.09
iii) Transactional only 1139 63.8% 6.3 9.7 3.5 8.17 8.06 -0.113 8.648 8.88 0.24
Total 1784 100.0% 6.2 9.6 3.4 8.25 8.11 -0.143 8.760 8.90 0.14
Note: L-Firms are those that went into default in the period 2008:q3-2010:q1, H-Firms are the remaining ones.

54

Table 2 Effect of Bank-firm relationship on the marginal probability of a firms default



Dependent variable: P(default
k
=1)
(I)
Baseline
equation
(II)
Firm specific
characteristics
(III)
Alternative
Weight
Coef. Sig. Coef. Sig. Coef. Sig.

T-share (in value) 0.0032 *** 0.0029 ***
(0.0008) (0.0007)
T-share (number of banks) 0.0028 ***
(0.0007)

0.0051 *** 0.0051 ***

(0.0005) (0.0001)
LTD -0.0002 -0.0002
(0.0018) (0.0018)
Small firm -0.0021 -0.0021
(0.0034) (0.0034)
CREDIT_HISTORY -0.0002 *** -0.0001 ***

(0.0000) (0.0000)
Bank fixed effects Yes Yes Yes
Industry-province dummies Yes Yes Yes
Number of obs. 72,489 72,489 72,489
Pseudo R
2
0.1273 0.1395 0.1397
The models estimate the marginal probability for a firm k to go into default in the period 2008:q3-
2010:q1. All explanatory variables are evaluated at 2008:q2, prior Lehman's default. The variable
T- Share indicates the proportion of loans that firm k has borrowed from a transactional bank. We
report the share both in loan value and in terms of number of T-banks. Parameter estimates are
reported with robust standard errors in brackets (cluster at individual bank level). The symbols *,
**, and *** represent significance levels of 10%, 5%, and 1% respectively. Coefficients for
industry-province dummies and bank fixed effects are not reported.

55

Table 3 T-banking and R-banking in good times and bad times
Variables

Interest rate
good time
(2007:q2)
(I)
Interest rate
bad time
(2010:q1)
(II)
Log Loans
good time
(2007:q2)
(III)
Log Loans
bad time
(2010:q1)
(IV)

T-Bank -0.0805*** 0.1227*** -0.2753*** -0.3129***
(0.0174) (0.0210) (0.0123) (0.0110)

Bank fixed effects yes yes yes yes
Firm fixed effects yes yes yes yes
Number of obs. 184,859 184,859 184,859 184,859
Adjusted R-squared 0.529 0.585 0.426 0.473
Notes: The models in column (I) and (III) are estimated in 2007:q2; those in columns (II) and (IV) in
2010:q1. The dummy T-Bank takes the value of 1 if the loan is granted by a transactional bank. The
coefficients represent the difference relative to relationship banking (R-banks). Parameter estimates
are reported with robust standard errors in brackets (cluster at individual firm level). The symbols *,
**, and *** represent significance levels of 10%, 5%, and 1% respectively. Coefficients for fixed
effects are not reported.
56

Table 4 Comparing T-banking, R-banking and firms quality
Variables

Interest rate
good time
(2007:q2)
(I)
Interest rate
bad time
(2010:q1)
(II)
Log Loans
good time
(2007:q2)
(III)
Log Loans
bad time
(2010:q1)
(IV)

T-Bank -0.3309*** -0.3977*** 0.0795* 0.1023**
(0.0604) (0.0737) (0.0393) (0.0413)
R-Bank*Z 0.3479*** 0.5016*** 0.1036*** 0.1329***
(0.0148) (0.0178) (0.0115) (0.0096)
T-Bank*Z 0.4238*** 0.7076*** 0.0575*** 0.0577***
(0.0119) (0.0151) (0.0092) (0.0062)
US>GR 0.8825*** 1.5181*** 0.6887*** 0.5667***
(0.0193) (0.0192) (0.0093) (0.0075)
LTD -0.3697*** -0.3760*** -0.0603* -0.0796***
(0.0453) (0.0561) (0.0330) (0.0213)
Small firm -0.0854 0.2037 -0.3993*** -0.4688***
(0.2295) (0.2463) (0.0968) (0.0784)
CREDIT_HISTORY -0.0475*** -0.0619*** 0.0460*** 0.0404***
(0.0020) (0.0023) (0.0013) (0.0009)
Bank fixed effects yes yes yes yes
Firm fixed effects no no no no
Industry-province dummies yes yes yes yes
Number of obs. 184,859 184,859 184,859 184,859
Adjusted R-squared 0.1776 0.2065 0.0865 0.0857
Notes: The models in column (I) and (III) are estimated in 2007:q2; those in columns (II) and
(IV) in 2010:q1. The dummy T-Bank takes the value of 1 if the loan is granted by a transactional
bank. The coefficients represent the difference relative to relationship banking (R-banks).
Parameter estimates are reported with robust standard errors in brackets (cluster at individual
bank level). The symbols *, **, and *** represent significance levels of 10%, 5%, and 1%
respectively. Coefficients for industry-province dummies and fixed effects are not reported.

57

Table 5 Capital endowment and bank type
Variables

Baseline
model

(I)
Bank-specific
characteristics

(II)
Firm-specific
characteristics

(III)
Financially
constrained
firms
(IV)

T-share -3.839*** -3.276** -3.091** -3.203**
(0.890) (1.301) (1.267) (1.265)
Bank size -0.040 0.090 0.092
(0.280) (0.268) (0.264)
Bank liquidity ratio -0.009 -0.011 -0.003
(0.016) (0.019) (0.017)
Retail ratio 0.049*** 0.031* 0.029*
(0.017) (0.017) (0.017)
Proportion of small firms in the
banks credit portfolio 6.169 5.972
(4.248) (4.115)
Proportion of LTD firms in the
banks credit portfolio -2.422 -2.141
(3.723) (3.712)
Average Z-score of the banks
credit portfolio -1.611 -1.337
(2.468) (2.563)
Proportion of financially
constrained firms (US>GR) 5.392
(6.603)
Bank zone dummies yes yes yes yes


Number of obs. 179 179 179 179
Adjusted R-squared 0.130 0.185 0.217 0.218
Notes: The dependent variable is the regulatory capital/risk-weighted asset ratio at 2008:q2 prior
to Lehmans default. The variable T-share represents the proportion of transactional loans (in
value) for bank j. It takes the value from 0 (pure R-bank) to 1 (pure T-bank). All bank-specific
characteristics and credit portfolio characteristic are at 2008:q1. Parameter estimates are reported
with robust standard errors in brackets (cluster at individual bank level). The symbols *, **, and
*** represent significance levels of 10%, 5%, and 1% respectively. Coefficients for bank zone
dummies are not reported.

58

Table 6 Lending relationship and bank-capital
Variables
Interest rate
good time
(2007:q2)
(I)
Interest rate
bad time
(2010:q1)
(II)
Log Loans
good time
(2007:q2)
(III)
Log Loans
bad time
(2010:q1)
(IV)

T-Bank -0.0792** 0.1940** -0.1625*** -0.2208***
(0.0402) (0.0734) (0.0282) (0.0289)
CAP 0.0096 -0.0426*** -0.0112 0.0113**
(0.0185) (0.0123) (0.0086) (0.0052)
US>GR 0.1881*** 0.1611*** 0.5315*** 0.1403***
(0.0228) (0.0430) (0.0174) (0.0193)
MUTUAL -0.7812*** -1.0057*** 0.0573 0.0569
(0.1284) (0.1066) (0.0378) (0.0523)
Bank group and rescue dummies yes yes yes yes
Bank zone dummies yes yes yes yes
Firm fixed effects yes yes yes yes

Number of obs. 184,859 184,859 184,859 184,859
Adjusted R-squared 0.4856 0.5433 0.4161 0.4530
Notes: The models in column (I) and (III) are estimated in 2007:q2; those in columns (II) and (IV) in
2010:q1. The dummy T-Bank takes the value of 1 if the loan is granted by a transactional bank. The
coefficients represent the difference relative to relationship banking (R-banks). Parameter estimates are
reported with robust standard errors in brackets (cluster at individual bank group level). The symbols *, **,
and *** represent significance levels of 10%, 5%, and 1% respectively. Coefficients for dummies and firm
fixed effects are not reported.


59

Table C1 Effect of Bank-firm relationship on the marginal probability of a firms
default. Including Main bank dummy and its interaction with T-share.



Dependent variable: P(default
k
=1)
(I)
Baseline equation
(II)
Firm specific
characteristics
(III)
Alternative
Weight
Coef. Sig. Coef. Sig. Coef. Sig.

T-share (in value) 0.0042 *** 0.0036 ***
(0.0011) (0.0009)
T-share (number of banks) 0.0036 ***
(0.0015)
Maxsh
-0.0123 *** -0.0108 *** -0.0106 ***

(0.0022) (0.0018) (0.0019)


Maxsh*T-share(in value) -0.0041 * -0.0033 *
(0.0023) (0.0019)
Maxsh*T-share(number of banks) -0.0035 *
(0.0020)

0.0048 *** 0.0048 ***

(0.0004) (0.0004)
LTD -0.0006 -0.0006
(0.0017) (0.0017)
Small firm -0.0020 -0.0020
(0.0032) (0.0032)
CREDIT_HISTORY -0.0002 *** -0.0002 ***

(0.0001) (0.0001)
Bank fixed effects Yes Yes Yes
Industry-province dummies Yes Yes Yes
Number of obs. 72,489 72,489 72,489
Pseudo R
2
0.0782 0.1190 0.1190
The models estimate the marginal probability for a firm k to go into default in the period 2008:q3-
2010:q1. All explanatory variables are evaluated at 2008:q2, prior Lehman's default. The variable T-
Share indicates the proportion of loans that firm k has borrowed from a transactional bank. We report
the share both in loan value and in terms of number of T-banks. The variable Maxsh indicates the
highest share of lending that is granted by the main bank. Parameter estimates are reported with robust
standard errors in brackets (cluster at individual bank level). The symbols *, **, and *** represent
significance levels of 10%, 5%, and 1% respectively. Coefficients for industry-province dummies and
bank fixed effects are not reported.

60

Table C2 T-banking and R-banking in good times and bad times. Including Main bank
dummy and its interaction with T-bank.
Variables

Interest rate
good time
(2007:q2)
(I)
Interest rate
bad time
(2010:q1)
(II)
Log Loans
good time
(2007:q2)
(III)
Log Loans
bad time
(2010:q1)
(IV)

T-Bank -0.0896*** 0.1086*** -0.1504*** -0.2067***
(0.0201) (0.0243) (0.0121) (0.0116)
Main -0.0969*** -0.1705*** 1.1652*** 0.8594***
(0.0232) (0.0281) (0.0130) (0.0130)
Main*T-Bank -0.0080 -0.0233 0.0325 0.0125
(0.0301) (0.0361) (0.0366) (0.0164)

Bank fixed effects yes yes yes yes
Firm fixed effects yes yes yes yes
Number of obs. 184,859 184,859 184,859 184,859
Adjusted R-squared 0.529 0.586 0.585 0.570
Notes: The models in column (I) and (III) are estimated in 2007:q2; those in columns (II) and (IV) in
2010:q1. The dummy T-Bank takes the value of 1 if the loan is granted by a transactional bank. The
coefficients represent the difference relative to relationship banking (R-banks). The dummy Main is
equal to one if that bank grants the highest share of lending to that firm. Parameter estimates are
reported with robust standard errors in brackets (cluster at individual firm level). The symbols *, **,
and *** represent significance levels of 10%, 5%, and 1% respectively. Coefficients for fixed effects
are not reported.

61

Table C3 Effect of Bank-firm relationship on the marginal probability of a firms
default. Changing relationship lending definition from province to region.



Dependent variable: P(default
k
=1)
(I)
Baseline
equation
(II)
Firm specific
characteristics
(III)
Alternative
Weight
Coef. Sig. Coef. Sig. Coef. Sig.

T-share (in value) 0.0024 *** 0.0024 ***
(0.0007) (0.0007)
T-share (number of banks) 0.0034 ***
(0.0007)

0.0051 *** 0.0051 ***

(0.0004) (0.0004)
LTD -0.0002 -0.0002
(0.0018) (0.0018)
Small firm -0.0020 -0.0018
(0.0034) (0.0035)
CREDIT_HISTORY -0.0001 ** -0.0002 **

(0.0000) (0.0000)
Bank fixed effects Yes Yes Yes
Industry-province dummies Yes Yes Yes
Number of obs. 72,489 72,489 72,489
Pseudo R
2
0.0612 0.1004 0.1003
The models estimate the marginal probability for a firm k to go into default in the period 2008:q3-
2010:q1. All explanatory variables are evaluated at 2008:q2, prior Lehman's default. The variable
T- Share indicates the proportion of loans that firm k has borrowed from a transactional bank. We
report the share both in loan value and in terms of number of T-banks. Parameter estimates are
reported with robust standard errors in brackets (cluster at individual bank level). The symbols *,
**, and *** represent significance levels of 10%, 5%, and 1% respectively. Coefficients for
industry-province dummies and bank fixed effects are not reported.

62

Table C4 T-banking and R-banking in good times and bad times. Changing
relationship lending definition from province to region.
Variables

Interest rate
good time
(2007:q2)
(I)
Interest rate
bad time
(2010:q1)
(II)
Log Loans
good time
(2007:q2)
(III)
Log Loans
bad time
(2010:q1)
(IV)

T-Bank -0.0748*** 0.1038*** -0.2428*** -0.256***
(0.0182) (0.0217) (0.0123) (0.0110)

Bank fixed effects yes yes yes yes
Firm fixed effects yes yes yes yes
Number of obs. 184,859 184,859 184,859 184,859
Adjusted R-squared 0.529 0.585 0.426 0.472
Notes: The models in column (I) and (III) are estimated in 2007:q2; those in columns (II) and (IV) in
2010:q1. The dummy T-Bank takes the value of 1 if the loan is granted by a transactional bank. The
coefficients represent the difference relative to relationship banking (R-banks). Parameter estimates
are reported with robust standard errors in brackets (cluster at individual firm level). The symbols *,
**, and *** represent significance levels of 10%, 5%, and 1% respectively. Coefficients for fixed
effects are not reported.

63

Table C5 Effect of Bank-firm relationship on the marginal probability of a firms
default. All foreign banks subsidiaries are T-banks.



Dependent variable: P(default
k
=1)
(I)
Baseline
equation
(II)
Firm specific
characteristics
(III)
Alternative
Weight
Coef. Sig. Coef. Sig. Coef. Sig.

T-share (in value) 0.0031 *** 0.0027 ***
(0.0009) (0.0007)
T-share (number of banks) 0.0027 ***
(0.0007)

0.0051 *** 0.0051 ***

(0.0004) (0.0004)
LTD -0.0002 -0.0002
(0.0018) (0.0018)
Small firm -0.0021 -0.0021
(0.0034) (0.0034)
CREDIT_HISTORY -0.0002 ** -0.0002 **

(0.0000) (0.0000)
Bank fixed effects Yes Yes Yes
Industry-province dummies Yes Yes Yes
Number of obs. 72,489 72,489 72,489
Pseudo R
2
0.0600 0.0994 0.0994
The models estimate the marginal probability for a firm k to go into default in the period 2008:q3-
2010:q1. All explanatory variables are evaluated at 2008:q2, prior Lehman's default. The variable
T- Share indicates the proportion of loans that firm k has borrowed from a transactional bank. We
report the share both in loan value and in terms of number of T-banks. Parameter estimates are
reported with robust standard errors in brackets (cluster at individual bank level). The symbols *,
**, and *** represent significance levels of 10%, 5%, and 1% respectively. Coefficients for
industry-province dummies and bank fixed effects are not reported.

64

Table C6 T-banking and R-banking in good times and bad times. All foreign
banks subsidiaries are T-banks.
Variables

Interest rate
good time
(2007:q2)
(I)
Interest rate
bad time
(2010:q1)
(II)
Log Loans
good time
(2007:q2)
(III)
Log Loans
bad time
(2010:q1)
(IV)

T-Bank -0.0844*** 0.1030*** -0.2737*** -0.2970***
(0.0180) (0.0218) (0.0128) (0.0115)

Bank fixed effects yes yes yes yes
Firm fixed effects yes yes yes yes
Number of obs. 184,859 184,859 184,859 184,859
Adjusted R-squared 0.529 0.585 0.426 0.473
Notes: The models in column (I) and (III) are estimated in 2007:q2; those in columns (II) and (IV) in
2010:q1. The dummy T-Bank takes the value of 1 if the loan is granted by a transactional bank. The
coefficients represent the difference relative to relationship banking (R-banks). Parameter estimates
are reported with robust standard errors in brackets (cluster at individual firm level). The symbols *,
**, and *** represent significance levels of 10%, 5%, and 1% respectively. Coefficients for fixed
effects are not reported.

65

Table C7 Effect of Bank-firm relationship on the marginal probability of a firms
default. New Firms.



Dependent variable: P(default
k
=1)
(I)
Baseline
equation
(II)
Firm specific
characteristics
(III)
Alternative
Weight
Coef. Sig. Coef. Sig. Coef. Sig.

T-share (in value) 0.0120 ** 0.0073 **
(0.0054) (0.0035)
T-share (number of banks) 0.0067 **
(0.0033)

0.0036 0.0036

(0.0024) (0.0024)
LTD -0.0018 -0.0018
(0.0073) (0.0073)
Small firm 0.0070 0.0070
(0.0025) (0.0025)
CREDIT_HISTORY 0.0033 0.0033

(0.0035) (0.0035)
Bank fixed effects Yes Yes Yes
Industry-province dummies Yes Yes Yes
Number of obs. 5,866 5,866 5,866
Pseudo R
2
0.0596 0.1470 0.1470
The models estimate the marginal probability for a firm k to go into default in the period 2008:q3-
2010:q1. All explanatory variables are evaluated at 2008:q2, prior Lehman's default. The variable
T- Share indicates the proportion of loans that firm k has borrowed from a transactional bank. We
report the share both in loan value and in terms of number of T-banks. Parameter estimates are
reported with robust standard errors in brackets (cluster at individual bank level). The symbols *,
**, and *** represent significance levels of 10%, 5%, and 1% respectively. Coefficients for
industry-province dummies and bank fixed effects are not reported.

66

Table C8 T-banking and R-banking in good times and bad times. New Firms.
Variables

Interest rate
good time
(2007:q2)
(I)
Interest rate
bad time
(2010:q1)
(II)
Log Loans
good time
(2007:q2)
(III)
Log Loans
bad time
(2010:q1)
(IV)

T-Bank -0.0844*** 0.1030*** -0.2737*** -0.2970***
(0.0180) (0.0218) (0.0128) (0.0115)

Bank fixed effects yes yes yes yes
Firm fixed effects yes yes yes yes
Number of obs. 5,866 5,866 5,866 5,866
Adjusted R-squared 0.529 0.585 0.426 0.473
Notes: The models in column (I) and (III) are estimated in 2007:q2; those in columns (II) and (IV) in
2010:q1. The dummy T-Bank takes the value of 1 if the loan is granted by a transactional bank. The
coefficients represent the difference relative to relationship banking (R-banks). Parameter estimates
are reported with robust standard errors in brackets (cluster at individual firm level). The symbols *,
**, and *** represent significance levels of 10%, 5%, and 1% respectively. Coefficients for fixed
effects are not reported.